id
stringlengths
9
16
title
stringlengths
4
278
categories
stringlengths
5
104
abstract
stringlengths
6
4.09k
2502.14829
Measuring Faithfulness of Chains of Thought by Unlearning Reasoning Steps
cs.CL
When prompted to think step-by-step, language models (LMs) produce a chain of thought (CoT), a sequence of reasoning steps that the model supposedly used to produce its prediction. However, despite much work on CoT prompting, it is unclear if CoT reasoning is faithful to the models' parameteric beliefs. We introduce a framework for measuring parametric faithfulness of generated reasoning, and propose Faithfulness by Unlearning Reasoning steps (FUR), an instance of this framework. FUR erases information contained in reasoning steps from model parameters. We perform experiments unlearning CoTs of four LMs prompted on four multi-choice question answering (MCQA) datasets. Our experiments show that FUR is frequently able to change the underlying models' prediction by unlearning key steps, indicating when a CoT is parametrically faithful. Further analysis shows that CoTs generated by models post-unlearning support different answers, hinting at a deeper effect of unlearning. Importantly, CoT steps identified as important by FUR do not align well with human notions of plausbility, emphasizing the need for specialized alignment
2502.14830
Middle-Layer Representation Alignment for Cross-Lingual Transfer in Fine-Tuned LLMs
cs.CL cs.AI
While large language models demonstrate remarkable capabilities at task-specific applications through fine-tuning, extending these benefits across diverse languages is essential for broad accessibility. However, effective cross-lingual transfer is hindered by LLM performance gaps across languages and the scarcity of fine-tuning data in many languages. Through analysis of LLM internal representations from over 1,000+ language pairs, we discover that middle layers exhibit the strongest potential for cross-lingual alignment. Building on this finding, we propose a middle-layer alignment objective integrated into task-specific training. Our experiments on slot filling, machine translation, and structured text generation show consistent improvements in cross-lingual transfer, especially to lower-resource languages. The method is robust to the choice of alignment languages and generalizes to languages unseen during alignment. Furthermore, we show that separately trained alignment modules can be merged with existing task-specific modules, improving cross-lingual capabilities without full re-training. Our code is publicly available (https://github.com/dannigt/mid-align).
2502.14831
Improving the Diffusability of Autoencoders
cs.CV cs.AI cs.LG
Latent diffusion models have emerged as the leading approach for generating high-quality images and videos, utilizing compressed latent representations to reduce the computational burden of the diffusion process. While recent advancements have primarily focused on scaling diffusion backbones and improving autoencoder reconstruction quality, the interaction between these components has received comparatively less attention. In this work, we perform a spectral analysis of modern autoencoders and identify inordinate high-frequency components in their latent spaces, which are especially pronounced in the autoencoders with a large bottleneck channel size. We hypothesize that this high-frequency component interferes with the coarse-to-fine nature of the diffusion synthesis process and hinders the generation quality. To mitigate the issue, we propose scale equivariance: a simple regularization strategy that aligns latent and RGB spaces across frequencies by enforcing scale equivariance in the decoder. It requires minimal code changes and only up to 20K autoencoder fine-tuning steps, yet significantly improves generation quality, reducing FID by 19% for image generation on ImageNet-1K 256x256 and FVD by at least 44% for video generation on Kinetics-700 17x256x256.
2502.14833
Probabilistic Robustness in Deep Learning: A Concise yet Comprehensive Guide
cs.LG
Deep learning (DL) has demonstrated significant potential across various safety-critical applications, yet ensuring its robustness remains a key challenge. While adversarial robustness has been extensively studied in worst-case scenarios, probabilistic robustness (PR) offers a more practical perspective by quantifying the likelihood of failures under stochastic perturbations. This paper provides a concise yet comprehensive overview of PR, covering its formal definitions, evaluation and enhancement methods. We introduce a reformulated ''min-max'' optimisation framework for adversarial training specifically designed to improve PR. Furthermore, we explore the integration of PR verification evidence into system-level safety assurance, addressing challenges in translating DL model-level robustness to system-level claims. Finally, we highlight open research questions, including benchmarking PR evaluation methods, extending PR to generative AI tasks, and developing rigorous methodologies and case studies for system-level integration.
2502.14834
LongWriter-V: Enabling Ultra-Long and High-Fidelity Generation in Vision-Language Models
cs.CV cs.AI cs.CL
Existing Large Vision-Language Models (LVLMs) can process inputs with context lengths up to 128k visual and text tokens, yet they struggle to generate coherent outputs beyond 1,000 words. We find that the primary limitation is the absence of long output examples during supervised fine-tuning (SFT). To tackle this issue, we introduce LongWriter-V-22k, a SFT dataset comprising 22,158 examples, each with multiple input images, an instruction, and corresponding outputs ranging from 0 to 10,000 words. Moreover, to achieve long outputs that maintain high-fidelity to the input images, we employ Direct Preference Optimization (DPO) to the SFT model. Given the high cost of collecting human feedback for lengthy outputs (e.g., 3,000 words), we propose IterDPO, which breaks long outputs into segments and uses iterative corrections to form preference pairs with the original outputs. Additionally, we develop MMLongBench-Write, a benchmark featuring six tasks to evaluate the long-generation capabilities of VLMs. Our 7B parameter model, trained with LongWriter-V-22k and IterDPO, achieves impressive performance on this benchmark, outperforming larger proprietary models like GPT-4o. Code and data: https://github.com/THU-KEG/LongWriter-V
2502.14837
Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs
cs.CL cs.AI
Multi-head Latent Attention (MLA) is an innovative architecture proposed by DeepSeek, designed to ensure efficient and economical inference by significantly compressing the Key-Value (KV) cache into a latent vector. Compared to MLA, standard LLMs employing Multi-Head Attention (MHA) and its variants such as Grouped-Query Attention (GQA) exhibit significant cost disadvantages. Enabling well-trained LLMs (e.g., Llama) to rapidly adapt to MLA without pre-training from scratch is both meaningful and challenging. This paper proposes the first data-efficient fine-tuning method for transitioning from MHA to MLA (MHA2MLA), which includes two key components: for partial-RoPE, we remove RoPE from dimensions of queries and keys that contribute less to the attention scores, for low-rank approximation, we introduce joint SVD approximations based on the pre-trained parameters of keys and values. These carefully designed strategies enable MHA2MLA to recover performance using only a small fraction (0.3% to 0.6%) of the data, significantly reducing inference costs while seamlessly integrating with compression techniques such as KV cache quantization. For example, the KV cache size of Llama2-7B is reduced by 92.19%, with only a 0.5% drop in LongBench performance.
2502.14838
Revealing and Mitigating Over-Attention in Knowledge Editing
cs.CL cs.AI
Large Language Models have demonstrated superior performance across a wide range of tasks, but they still exhibit undesirable errors due to incorrect knowledge learned from the training data. To avoid this, knowledge editing methods emerged to precisely edit the specific model knowledge via efficiently modifying a very small percentage of parameters. % However, those methods can lead to the problem of Specificity Failure: when the content related to the edited knowledge occurs in the context, it can inadvertently corrupt other pre-existing knowledge. However, those methods can lead to the problem of Specificity Failure, where the existing knowledge and capabilities are severely degraded due to editing. Our preliminary indicates that Specificity Failure primarily stems from the model's attention heads assigning excessive attention scores to entities related to the edited knowledge, thereby unduly focusing on specific snippets within the context, which we denote as the Attention Drift phenomenon. To mitigate such Attention Drift issue, we introduce a simple yet effective method Selective Attention Drift Restriction}(SADR), which introduces an additional regularization term during the knowledge editing process to restrict changes in the attention weight distribution, thereby preventing undue focus on the edited entity. Experiments on five frequently used strong LLMs demonstrate the effectiveness of our method, where SADR can significantly mitigate Specificity Failure in the predominant knowledge editing tasks.
2502.14840
Spatial Distribution-Shift Aware Knowledge-Guided Machine Learning
cs.LG
Given inputs of diverse soil characteristics and climate data gathered from various regions, we aimed to build a model to predict accurate land emissions. The problem is important since accurate quantification of the carbon cycle in agroecosystems is crucial for mitigating climate change and ensuring sustainable food production. Predicting accurate land emissions is challenging since calibrating the heterogeneous nature of soil properties, moisture, and environmental conditions is hard at decision-relevant scales. Traditional approaches do not adequately estimate land emissions due to location-independent parameters failing to leverage the spatial heterogeneity and also require large datasets. To overcome these limitations, we proposed Spatial Distribution-Shift Aware Knowledge-Guided Machine Learning (SDSA-KGML), which leverages location-dependent parameters that account for significant spatial heterogeneity in soil moisture from multiple sites within the same region. Experimental results demonstrate that SDSA-KGML models achieve higher local accuracy for the specified states in the Midwest Region.
2502.14842
Generating $\pi$-Functional Molecules Using STGG+ with Active Learning
cs.LG
Generating novel molecules with out-of-distribution properties is a major challenge in molecular discovery. While supervised learning methods generate high-quality molecules similar to those in a dataset, they struggle to generalize to out-of-distribution properties. Reinforcement learning can explore new chemical spaces but often conducts 'reward-hacking' and generates non-synthesizable molecules. In this work, we address this problem by integrating a state-of-the-art supervised learning method, STGG+, in an active learning loop. Our approach iteratively generates, evaluates, and fine-tunes STGG+ to continuously expand its knowledge. We denote this approach STGG+AL. We apply STGG+AL to the design of organic $\pi$-functional materials, specifically two challenging tasks: 1) generating highly absorptive molecules characterized by high oscillator strength and 2) designing absorptive molecules with reasonable oscillator strength in the near-infrared (NIR) range. The generated molecules are validated and rationalized in-silico with time-dependent density functional theory. Our results demonstrate that our method is highly effective in generating novel molecules with high oscillator strength, contrary to existing methods such as reinforcement learning (RL) methods. We open-source our active-learning code along with our Conjugated-xTB dataset containing 2.9 million $\pi$-conjugated molecules and the function for approximating the oscillator strength and absorption wavelength (based on sTDA-xTB).
2502.14844
Dynamic Concepts Personalization from Single Videos
cs.GR cs.CV cs.LG
Personalizing generative text-to-image models has seen remarkable progress, but extending this personalization to text-to-video models presents unique challenges. Unlike static concepts, personalizing text-to-video models has the potential to capture dynamic concepts, i.e., entities defined not only by their appearance but also by their motion. In this paper, we introduce Set-and-Sequence, a novel framework for personalizing Diffusion Transformers (DiTs)-based generative video models with dynamic concepts. Our approach imposes a spatio-temporal weight space within an architecture that does not explicitly separate spatial and temporal features. This is achieved in two key stages. First, we fine-tune Low-Rank Adaptation (LoRA) layers using an unordered set of frames from the video to learn an identity LoRA basis that represents the appearance, free from temporal interference. In the second stage, with the identity LoRAs frozen, we augment their coefficients with Motion Residuals and fine-tune them on the full video sequence, capturing motion dynamics. Our Set-and-Sequence framework results in a spatio-temporal weight space that effectively embeds dynamic concepts into the video model's output domain, enabling unprecedented editability and compositionality while setting a new benchmark for personalizing dynamic concepts.
2502.14846
Scaling Text-Rich Image Understanding via Code-Guided Synthetic Multimodal Data Generation
cs.CV cs.CL
Reasoning about images with rich text, such as charts and documents, is a critical application of vision-language models (VLMs). However, VLMs often struggle in these domains due to the scarcity of diverse text-rich vision-language data. To address this challenge, we present CoSyn, a framework that leverages the coding capabilities of text-only large language models (LLMs) to automatically create synthetic text-rich multimodal data. Given input text describing a target domain (e.g., "nutrition fact labels"), CoSyn prompts an LLM to generate code (Python, HTML, LaTeX, etc.) for rendering synthetic images. With the underlying code as textual representations of the synthetic images, CoSyn can generate high-quality instruction-tuning data, again relying on a text-only LLM. Using CoSyn, we constructed a dataset comprising 400K images and 2.7M rows of vision-language instruction-tuning data. Comprehensive experiments on seven benchmarks demonstrate that models trained on our synthetic data achieve state-of-the-art performance among competitive open-source models, including Llama 3.2, and surpass proprietary models such as GPT-4V and Gemini 1.5 Flash. Furthermore, CoSyn can produce synthetic pointing data, enabling VLMs to ground information within input images, showcasing its potential for developing multimodal agents capable of acting in real-world environments.
2502.14848
GATE: Graph-based Adaptive Tool Evolution Across Diverse Tasks
cs.CL
Large Language Models (LLMs) have shown great promise in tool-making, yet existing frameworks often struggle to efficiently construct reliable toolsets and are limited to single-task settings. To address these challenges, we propose GATE (Graph-based Adaptive Tool Evolution), an adaptive framework that dynamically constructs and evolves a hierarchical graph of reusable tools across multiple scenarios. We evaluate GATE on open-ended tasks (Minecraft), agent-based tasks (TextCraft, DABench), and code generation tasks (MATH, Date, TabMWP). Our results show that GATE achieves up to 4.3x faster milestone completion in Minecraft compared to the previous SOTA, and provides an average improvement of 9.23% over existing tool-making methods in code generation tasks and 10.03% in agent tasks. GATE demonstrates the power of adaptive evolution, balancing tool quantity, complexity, and functionality while maintaining high efficiency. Code and data are available at \url{https://github.com/ayanami2003/GATE}.
2502.14853
On the $H$-property for Step-graphons: Residual Case
eess.SY cs.SY
We sample graphs $G_n$ on $n$ nodes from a step-graphon and evaluate the probability that $G_n$ has a Hamiltonian decomposition in the asymptotic regime as $n\to\infty$. It has recently been shown that for almost all step-graphons, this probability converges to either zero or one. In this paper, we focus on the class of step-graphons such that the zero-one property does not hold. We show in this case that the limit of the probability still exists and provide an explicit expression of it.
2502.14854
CLIPPER: Compression enables long-context synthetic data generation
cs.CL
LLM developers are increasingly reliant on synthetic data, but generating high-quality data for complex long-context reasoning tasks remains challenging. We introduce CLIPPER, a compression-based approach for generating synthetic data tailored to narrative claim verification - a task that requires reasoning over a book to verify a given claim. Instead of generating claims directly from the raw text of the book, which results in artifact-riddled claims, CLIPPER first compresses the book into chapter outlines and book summaries and then uses these intermediate representations to generate complex claims and corresponding chain-of-thoughts. Compared to naive approaches, CLIPPER produces claims that are more valid, grounded, and complex. Using CLIPPER, we construct a dataset of 19K synthetic book claims paired with their source texts and chain-of-thought reasoning, and use it to fine-tune three open-weight models. Our best model achieves breakthrough results on narrative claim verification (from 28% to 76% accuracy on our test set) and sets a new state-of-the-art for sub-10B models on the NoCha leaderboard. Further analysis shows that our models generate more detailed and grounded chain-of-thought reasoning while also improving performance on other narrative understanding tasks (e.g., NarrativeQA).
2502.14855
Prompt-to-Leaderboard
cs.LG cs.CL
Large language model (LLM) evaluations typically rely on aggregated metrics like accuracy or human preference, averaging across users and prompts. This averaging obscures user- and prompt-specific variations in model performance. To address this, we propose Prompt-to-Leaderboard (P2L), a method that produces leaderboards specific to a prompt. The core idea is to train an LLM taking natural language prompts as input to output a vector of Bradley-Terry coefficients which are then used to predict the human preference vote. The resulting prompt-dependent leaderboards allow for unsupervised task-specific evaluation, optimal routing of queries to models, personalization, and automated evaluation of model strengths and weaknesses. Data from Chatbot Arena suggest that P2L better captures the nuanced landscape of language model performance than the averaged leaderboard. Furthermore, our findings suggest that P2L's ability to produce prompt-specific evaluations follows a power law scaling similar to that observed in LLMs themselves. In January 2025, the router we trained based on this methodology achieved the \#1 spot in the Chatbot Arena leaderboard. Our code is available at this GitHub link: https://github.com/lmarena/p2l.
2502.14856
FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling
cs.CL cs.AI cs.LG
Speculative sampling has emerged as an important technique for accelerating the auto-regressive generation process of large language models (LLMs) by utilizing a draft-then-verify mechanism to produce multiple tokens per forward pass. While state-of-the-art speculative sampling methods use only a single layer and a language modeling (LM) head as the draft model to achieve impressive layer compression, their efficiency gains are substantially reduced for large-vocabulary LLMs, such as Llama-3-8B with a vocabulary of 128k tokens. To address this, we present FR-Spec, a frequency-ranked speculative sampling framework that optimizes draft candidate selection through vocabulary space compression. By constraining the draft search to a frequency-prioritized token subset, our method reduces LM Head computation overhead by 75% while ensuring the equivalence of the final output distribution. Experiments across multiple datasets demonstrate an average of 1.12$\times$ speedup over the state-of-the-art speculative sampling method EAGLE-2.
2502.14860
Aligning LLMs to Ask Good Questions A Case Study in Clinical Reasoning
cs.CL
Large language models (LLMs) often fail to ask effective questions under uncertainty, making them unreliable in domains where proactive information-gathering is essential for decisionmaking. We present ALFA, a framework that improves LLM question-asking by (i) decomposing the notion of a "good" question into a set of theory-grounded attributes (e.g., clarity, relevance), (ii) controllably synthesizing attribute-specific question variations, and (iii) aligning models via preference-based optimization to explicitly learn to ask better questions along these fine-grained attributes. Focusing on clinical reasoning as a case study, we introduce the MediQ-AskDocs dataset, composed of 17k real-world clinical interactions augmented with 80k attribute-specific preference pairs of follow-up questions, as well as a novel expert-annotated interactive healthcare QA task to evaluate question-asking abilities. Models aligned with ALFA reduce diagnostic errors by 56.6% on MediQ-AskDocs compared to SOTA instruction-tuned LLMs, with a question-level win-rate of 64.4% and strong generalizability. Our findings suggest that explicitly guiding question-asking with structured, fine-grained attributes offers a scalable path to improve LLMs, especially in expert application domains.
2502.14862
Interpretable Text Embeddings and Text Similarity Explanation: A Primer
cs.CL cs.AI cs.IR
Text embeddings and text embedding models are a backbone of many AI and NLP systems, particularly those involving search. However, interpretability challenges persist, especially in explaining obtained similarity scores, which is crucial for applications requiring transparency. In this paper, we give a structured overview of interpretability methods specializing in explaining those similarity scores, an emerging research area. We study the methods' individual ideas and techniques, evaluating their potential for improving interpretability of text embeddings and explaining predicted similarities.
2502.14864
Benchmarking Multimodal RAG through a Chart-based Document Question-Answering Generation Framework
cs.AI cs.CV
Multimodal Retrieval-Augmented Generation (MRAG) enhances reasoning capabilities by integrating external knowledge. However, existing benchmarks primarily focus on simple image-text interactions, overlooking complex visual formats like charts that are prevalent in real-world applications. In this work, we introduce a novel task, Chart-based MRAG, to address this limitation. To semi-automatically generate high-quality evaluation samples, we propose CHARt-based document question-answering GEneration (CHARGE), a framework that produces evaluation data through structured keypoint extraction, crossmodal verification, and keypoint-based generation. By combining CHARGE with expert validation, we construct Chart-MRAG Bench, a comprehensive benchmark for chart-based MRAG evaluation, featuring 4,738 question-answering pairs across 8 domains from real-world documents. Our evaluation reveals three critical limitations in current approaches: (1) unified multimodal embedding retrieval methods struggles in chart-based scenarios, (2) even with ground-truth retrieval, state-of-the-art MLLMs achieve only 58.19% Correctness and 73.87% Coverage scores, and (3) MLLMs demonstrate consistent text-over-visual modality bias during Chart-based MRAG reasoning. The CHARGE and Chart-MRAG Bench are released at https://github.com/Nomothings/CHARGE.git.
2502.14865
Time Travel: A Comprehensive Benchmark to Evaluate LMMs on Historical and Cultural Artifacts
cs.CV cs.LG
Understanding historical and cultural artifacts demands human expertise and advanced computational techniques, yet the process remains complex and time-intensive. While large multimodal models offer promising support, their evaluation and improvement require a standardized benchmark. To address this, we introduce TimeTravel, a benchmark of 10,250 expert-verified samples spanning 266 distinct cultures across 10 major historical regions. Designed for AI-driven analysis of manuscripts, artworks, inscriptions, and archaeological discoveries, TimeTravel provides a structured dataset and robust evaluation framework to assess AI models' capabilities in classification, interpretation, and historical comprehension. By integrating AI with historical research, TimeTravel fosters AI-powered tools for historians, archaeologists, researchers, and cultural tourists to extract valuable insights while ensuring technology contributes meaningfully to historical discovery and cultural heritage preservation. We evaluate contemporary AI models on TimeTravel, highlighting their strengths and identifying areas for improvement. Our goal is to establish AI as a reliable partner in preserving cultural heritage, ensuring that technological advancements contribute meaningfully to historical discovery. Our code is available at: \url{https://github.com/mbzuai-oryx/TimeTravel}.
2502.14866
LServe: Efficient Long-sequence LLM Serving with Unified Sparse Attention
cs.CL cs.AI cs.DC cs.LG cs.PF
Large language models (LLMs) have shown remarkable potential in processing long sequences, yet efficiently serving these long-context models remains challenging due to the quadratic computational complexity of attention in the prefilling stage and the large memory footprint of the KV cache in the decoding stage. To address these issues, we introduce LServe, an efficient system that accelerates long-sequence LLM serving via hybrid sparse attention. This method unifies different hardware-friendly, structured sparsity patterns for both prefilling and decoding attention into a single framework, where computations on less important tokens are skipped block-wise. LServe demonstrates the compatibility of static and dynamic sparsity in long-context LLM attention. This design enables multiplicative speedups by combining these optimizations. Specifically, we convert half of the attention heads to nearly free streaming heads in both the prefilling and decoding stages. Additionally, we find that only a constant number of KV pages is required to preserve long-context capabilities, irrespective of context length. We then design a hierarchical KV page selection policy that dynamically prunes KV pages based on query-centric similarity. On average, LServe accelerates LLM prefilling by up to 2.9x and decoding by 1.3-2.1x over vLLM, maintaining long-context accuracy. Code is released at https://github.com/mit-han-lab/omniserve.
adap-org/9807003
Development and Evolution of Neural Networks in an Artificial Chemistry
adap-org cs.NE nlin.AO q-bio.PE
We present a model of decentralized growth for Artificial Neural Networks (ANNs) inspired by the development and the physiology of real nervous systems. In this model, each individual artificial neuron is an autonomous unit whose behavior is determined only by the genetic information it harbors and local concentrations of substrates modeled by a simple artificial chemistry. Gene expression is manifested as axon and dendrite growth, cell division and differentiation, substrate production and cell stimulation. We demonstrate the model's power with a hand-written genome that leads to the growth of a simple network which performs classical conditioning. To evolve more complex structures, we implemented a platform-independent, asynchronous, distributed Genetic Algorithm (GA) that allows users to participate in evolutionary experiments via the World Wide Web.
adap-org/9903003
Evolution of genetic organization in digital organisms
adap-org cs.NE nlin.AO q-bio.PE
We examine the evolution of expression patterns and the organization of genetic information in populations of self-replicating digital organisms. Seeding the experiments with a linearly expressed ancestor, we witness the development of complex, parallel secondary expression patterns. Using principles from information theory, we demonstrate an evolutionary pressure towards overlapping expressions causing variation (and hence further evolution) to sharply drop. Finally, we compare the overlapping sections of dominant genomes to those portions which are singly expressed and observe a significant difference in the entropy of their encoding.
alg-geom/9608018
Rank Two Bundles on Algebraic Curves and Decoding of Goppa Codes
alg-geom cs.IT math.AG math.IT
We study a connection between two topics: Decoding of Goppa codes arising from an algebraic curve, and rank two extensions of certain line bundles on the curve.
astro-ph/0008307
Science User Scenarios for a Virtual Observatory Design Reference Mission: Science Requirements for Data Mining
astro-ph cs.DB cs.DL cs.IR
The knowledge discovery potential of the new large astronomical databases is vast. When these are used in conjunction with the rich legacy data archives, the opportunities for scientific discovery multiply rapidly. A Virtual Observatory (VO) framework will enable transparent and efficient access, search, retrieval, and visualization of data across multiple data repositories, which are generally heterogeneous and distributed. Aspects of data mining that apply to a variety of science user scenarios with a VO are reviewed. The development of a VO should address the data mining needs of various astronomical research constituencies. By way of example, two user scenarios are presented which invoke applications and linkages of data across the catalog and image domains in order to address specific astrophysics research problems. These illustrate a subset of the desired capabilities and power of the VO, and as such they represent potential components of a VO Design Reference Mission.
astro-ph/0010583
Data Mining in Astronomical Databases
astro-ph cs.DB cs.DL cs.IR
A Virtual Observatory (VO) will enable transparent and efficient access, search, retrieval, and visualization of data across multiple data repositories, which are generally heterogeneous and distributed. Aspects of data mining that apply to a variety of science user scenarios with a VO are reviewed.
astro-ph/0402591
Evolutionary design of photometric systems and its application to Gaia
astro-ph cs.NE stat.ML
Designing a photometric system to best fulfil a set of scientific goals is a complex task, demanding a compromise between conflicting requirements and subject to various constraints. A specific example is the determination of stellar astrophysical parameters (APs) - effective temperature, metallicity etc. - across a wide range of stellar types. I present a novel approach to this problem which makes minimal assumptions about the required filter system. By considering a filter system as a set of free parameters it may be designed by optimizing some figure-of-merit (FoM) with respect to these parameters. In the example considered, the FoM is a measure of how well the filter system can `separate' stars with different APs. This separation is vectorial in nature, in the sense that the local directions of AP variance are preferably mutually orthogonal to avoid AP degeneracy. The optimization is carried out with an evolutionary algorithm, which uses principles of evolutionary biology to search the parameter space. This model, HFD (Heuristic Filter Design), is applied to the design of photometric systems for the Gaia space astrometry mission. The optimized systems show a number of interesting features, not least the persistence of broad, overlapping filters. These HFD systems perform as least as well as other proposed systems for Gaia, although inadequacies remain in all. The principles underlying HFD are quite generic and may be applied to filter design for numerous other projects, such as the search for specific types of objects or photometric redshift determination.
astro-ph/0502164
Particle Swarm Optimization: An efficient method for tracing periodic orbits in 3D galactic potentials
astro-ph cs.NA cs.NE math.NA nlin.CD
We propose the Particle Swarm Optimization (PSO) as an alternative method for locating periodic orbits in a three--dimensional (3D) model of barred galaxies. We develop an appropriate scheme that transforms the problem of finding periodic orbits into the problem of detecting global minimizers of a function, which is defined on the Poincar\'{e} Surface of Section (PSS) of the Hamiltonian system. By combining the PSO method with deflection techniques, we succeeded in tracing systematically several periodic orbits of the system. The method succeeded in tracing the initial conditions of periodic orbits in cases where Newton iterative techniques had difficulties. In particular, we found families of 2D and 3D periodic orbits associated with the inner 8:1 to 12:1 resonances, between the radial 4:1 and corotation resonances of our 3D Ferrers bar model. The main advantages of the proposed algorithm is its simplicity, its ability to work using function values solely, as well as its ability to locate many periodic orbits per run at a given Jacobian constant.
astro-ph/0504006
Virtual Observatory: From Concept to Implementation
astro-ph cs.CE
We review the origins of the Virtual Observatory (VO) concept, and the current status of the efforts in this field. VO is the response of the astronomical community to the challenges posed by the modern massive and complex data sets. It is a framework in which information technology is harnessed to organize, maintain, and explore the rich information content of the exponentially growing data sets, and to enable a qualitatively new science to be done with them. VO will become a complete, open, distributed, web-based framework for astronomy of the early 21st century. A number of significant efforts worldwide are now striving to convert this vision into reality. The technological and methodological challenges posed by the information-rich astronomy are also common to many other fields. We see a fundamental change in the way all science is done, driven by the information technology revolution.
astro-ph/0506110
Galactic Gradients, Postbiological Evolution and the Apparent Failure of SETI
astro-ph cs.AI physics.soc-ph
Motivated by recent developments impacting our view of Fermi's paradox (absence of extraterrestrials and their manifestations from our past light cone), we suggest a reassessment of the problem itself, as well as of strategies employed by SETI projects so far. The need for such reevaluation is fueled not only by the failure of searches thus far, but also by great advances recently made in astrophysics, astrobiology, computer science and future studies, which have remained largely ignored in SETI practice. As an example of the new approach, we consider the effects of the observed metallicity and temperature gradients in the Milky Way on the spatial distribution of hypothetical advanced extraterrestrial intelligent communities. While, obviously, properties of such communities and their sociological and technological preferences are entirely unknown, we assume that (1) they operate in agreement with the known laws of physics, and (2) that at some point they typically become motivated by a meta-principle embodying the central role of information-processing; a prototype of the latter is the recently suggested Intelligence Principle of Steven J. Dick. There are specific conclusions of practical interest to be drawn from coupling of these reasonable assumptions with the astrophysical and astrochemical structure of the Galaxy. In particular, we suggest that the outer regions of the Galactic disk are most likely locations for advanced SETI targets, and that intelligent communities will tend to migrate outward through the Galaxy as their capacities of information-processing increase, for both thermodynamical and astrochemical reasons. This can also be regarded as a possible generalization of the Galactic Habitable Zone, concept currently much investigated in astrobiology.
astro-ph/0506308
Fast directional continuous spherical wavelet transform algorithms
astro-ph cs.IT math.IT
We describe the construction of a spherical wavelet analysis through the inverse stereographic projection of the Euclidean planar wavelet framework, introduced originally by Antoine and Vandergheynst and developed further by Wiaux et al. Fast algorithms for performing the directional continuous wavelet analysis on the unit sphere are presented. The fast directional algorithm, based on the fast spherical convolution algorithm developed by Wandelt and Gorski, provides a saving of O(sqrt(Npix)) over a direct quadrature implementation for Npix pixels on the sphere, and allows one to perform a directional spherical wavelet analysis of a 10^6 pixel map on a personal computer.
astro-ph/0605042
How accurate are the time delay estimates in gravitational lensing?
astro-ph cs.LG
We present a novel approach to estimate the time delay between light curves of multiple images in a gravitationally lensed system, based on Kernel methods in the context of machine learning. We perform various experiments with artificially generated irregularly-sampled data sets to study the effect of the various levels of noise and the presence of gaps of various size in the monitoring data. We compare the performance of our method with various other popular methods of estimating the time delay and conclude, from experiments with artificial data, that our method is least vulnerable to missing data and irregular sampling, within reasonable bounds of Gaussian noise. Thereafter, we use our method to determine the time delays between the two images of quasar Q0957+561 from radio monitoring data at 4 cm and 6 cm, and conclude that if only the observations at epochs common to both wavelengths are used, the time delay gives consistent estimates, which can be combined to yield 408\pm 12 days. The full 6 cm dataset, which covers a longer monitoring period, yields a value which is 10% larger, but this can be attributed to differences in sampling and missing data.
astro-ph/0609159
A directional continuous wavelet transform on the sphere
astro-ph cs.IT math.IT
A new construction of a directional continuous wavelet analysis on the sphere is derived herein. We adopt the harmonic scaling idea for the spherical dilation operator recently proposed by Sanz et al. but extend the analysis to a more general directional framework. Directional wavelets are a powerful extension that allow one to also probe oriented structure in the analysed function. Our spherical wavelet methodology has the advantage that all functions and operators are defined directly on the sphere. The construction of wavelets in our framework is demonstrated with an example.
astro-ph/0612688
Optimal filters on the sphere
astro-ph cs.IT math.IT
We derive optimal filters on the sphere in the context of detecting compact objects embedded in a stochastic background process. The matched filter and the scale adaptive filter are derived on the sphere in the most general setting, allowing for directional template profiles and filters. The performance and relative merits of the two optimal filters are discussed. The application of optimal filter theory on the sphere to the detection of compact objects is demonstrated on simulated mock data. A naive detection strategy is adopted, with an initial aim of illustrating the application of the new optimal filters derived on the sphere. Nevertheless, this simple object detection strategy is demonstrated to perform well, even a low signal-to-noise ratio. Code written to compute optimal filters on the sphere (S2FIL), to perform fast directional filtering on the sphere (FastCSWT) and to construct the simulated mock data (COMB) are all made publicly available from http://www.mrao.cam.ac.uk/~jdm57/
cmp-lg/9404001
An Alternative Conception of Tree-Adjoining Derivation
cmp-lg cs.CL
The precise formulation of derivation for tree-adjoining grammars has important ramifications for a wide variety of uses of the formalism, from syntactic analysis to semantic interpretation and statistical language modeling. We argue that the definition of tree-adjoining derivation must be reformulated in order to manifest the proper linguistic dependencies in derivations. The particular proposal is both precisely characterizable through a definition of TAG derivations as equivalence classes of ordered derivation trees, and computationally operational, by virtue of a compilation to linear indexed grammars together with an efficient algorithm for recognition and parsing according to the compiled grammar.
cmp-lg/9404002
Lessons from a Restricted Turing Test
cmp-lg cs.CL
We report on the recent Loebner prize competition inspired by Turing's test of intelligent behavior. The presentation covers the structure of the competition and the outcome of its first instantiation in an actual event, and an analysis of the purpose, design, and appropriateness of such a competition. We argue that the competition has no clear purpose, that its design prevents any useful outcome, and that such a competition is inappropriate given the current level of technology. We then speculate as to suitable alternatives to the Loebner prize.
cmp-lg/9404003
Restricting the Weak-Generative Capacity of Synchronous Tree-Adjoining Grammars
cmp-lg cs.CL
The formalism of synchronous tree-adjoining grammars, a variant of standard tree-adjoining grammars (TAG), was intended to allow the use of TAGs for language transduction in addition to language specification. In previous work, the definition of the transduction relation defined by a synchronous TAG was given by appeal to an iterative rewriting process. The rewriting definition of derivation is problematic in that it greatly extends the expressivity of the formalism and makes the design of parsing algorithms difficult if not impossible. We introduce a simple, natural definition of synchronous tree-adjoining derivation, based on isomorphisms between standard tree-adjoining derivations, that avoids the expressivity and implementability problems of the original rewriting definition. The decrease in expressivity, which would otherwise make the method unusable, is offset by the incorporation of an alternative definition of standard tree-adjoining derivation, previously proposed for completely separate reasons, thereby making it practical to entertain using the natural definition of synchronous derivation. Nonetheless, some remaining problematic cases call for yet more flexibility in the definition; the isomorphism requirement may have to be relaxed. It remains for future research to tune the exact requirements on the allowable mappings.
cmp-lg/9404004
An Empirically Motivated Reinterpretation of Dependency Grammar
cmp-lg cs.CL
Dependency grammar is usually interpreted as equivalent to a strict form of X--bar theory that forbids the stacking of nodes of the same bar level (e.g., N' immediately dominating N' with the same head). But adequate accounts of _one_--anaphora and of the semantics of multiple modifiers require such stacking and accordingly argue against dependency grammar. Dependency grammar can be salvaged by reinterpreting its claims about phrase structure, so that modifiers map onto binary--branching X--bar trees rather than ``flat'' ones.
cmp-lg/9404005
Memoization in Constraint Logic Programming
cmp-lg cs.CL
This paper shows how to apply memoization (caching of subgoals and associated answer substitutions) in a constraint logic programming setting. The research is is motivated by the desire to apply constraint logic programming (CLP) to problems in natural language processing that involve (constraint) interleaving or coroutining, such as GB and HPSG parsing.
cmp-lg/9404006
SPANISH 1992 (S92): corpus-based analysis of present-day Spanish for medical purposes
cmp-lg cs.CL
S92 research was begun in 1987 to analyze word frequencies in present-day Spanish for making speech pathology evaluation tools. 500 2,000-word samples of children, adolescents and adults' language were input between 1988-1991, calculations done in 1992; statistical and Lewandowski analyses were carried out in 1993.
cmp-lg/9404007
Constraint-Based Categorial Grammar
cmp-lg cs.CL
We propose a generalization of Categorial Grammar in which lexical categories are defined by means of recursive constraints. In particular, the introduction of relational constraints allows one to capture the effects of (recursive) lexical rules in a computationally attractive manner. We illustrate the linguistic merits of the new approach by showing how it accounts for the syntax of Dutch cross-serial dependencies and the position and scope of adjuncts in such constructions. Delayed evaluation is used to process grammars containing recursive constraints.
cmp-lg/9404008
Principles and Implementation of Deductive Parsing
cmp-lg cs.CL
We present a system for generating parsers based directly on the metaphor of parsing as deduction. Parsing algorithms can be represented directly as deduction systems, and a single deduction engine can interpret such deduction systems so as to implement the corresponding parser. The method generalizes easily to parsers for augmented phrase structure formalisms, such as definite-clause grammars and other logic grammar formalisms, and has been used for rapid prototyping of parsing algorithms for a variety of formalisms including variants of tree-adjoining grammars, categorial grammars, and lexicalized context-free grammars.
cmp-lg/9404009
A Deductive Account of Quantification in LFG
cmp-lg cs.CL
The relationship between Lexical-Functional Grammar (LFG) functional structures (f-structures) for sentences and their semantic interpretations can be expressed directly in a fragment of linear logic in a way that explains correctly the constrained interactions between quantifier scope ambiguity and bound anaphora. The use of a deductive framework to account for the compositional properties of quantifying expressions in natural language obviates the need for additional mechanisms, such as Cooper storage, to represent the different scopes that a quantifier might take. Instead, the semantic contribution of a quantifier is recorded as an ordinary logical formula, one whose use in a proof will establish the scope of the quantifier. The properties of linear logic ensure that each quantifier is scoped exactly once. Our analysis of quantifier scope can be seen as a recasting of Pereira's analysis (Pereira, 1991), which was expressed in higher-order intuitionistic logic. But our use of LFG and linear logic provides a much more direct and computationally more flexible interpretation mechanism for at least the same range of phenomena. We have developed a preliminary Prolog implementation of the linear deductions described in this work.
cmp-lg/9404010
Intensional Verbs Without Type-Raising or Lexical Ambiguity
cmp-lg cs.CL
We present an analysis of the semantic interpretation of intensional verbs such as seek that allows them to take direct objects of either individual or quantifier type, producing both de dicto and de re readings in the quantifier case, all without needing to stipulate type-raising or quantifying-in rules. This simple account follows directly from our use of logical deduction in linear logic to express the relationship between syntactic structures and meanings. While our analysis resembles current categorial approaches in important ways, it differs from them in allowing the greater type flexibility of categorial semantics while maintaining a precise connection to syntax. As a result, we are able to provide derivations for certain readings of sentences with intensional verbs and complex direct objects that are not derivable in current purely categorial accounts of the syntax-semantics interface. The analysis forms a part of our ongoing work on semantic interpretation within the framework of Lexical-Functional Grammar.
cmp-lg/9404011
Adjuncts and the Processing of Lexical Rules
cmp-lg cs.CL
The standard HPSG analysis of Germanic verb clusters can not explain the observed narrow-scope readings of adjuncts in such verb clusters. We present an extension of the HPSG analysis that accounts for the systematic ambiguity of the scope of adjuncts in verb cluster constructions, by treating adjuncts as members of the subcat list. The extension uses powerful recursive lexical rules, implemented as complex constraints. We show how `delayed evaluation' techniques from constraint-logic programming can be used to process such lexical rules.
cmp-lg/9405001
Similarity-Based Estimation of Word Cooccurrence Probabilities
cmp-lg cs.CL
In many applications of natural language processing it is necessary to determine the likelihood of a given word combination. For example, a speech recognizer may need to determine which of the two word combinations ``eat a peach'' and ``eat a beach'' is more likely. Statistical NLP methods determine the likelihood of a word combination according to its frequency in a training corpus. However, the nature of language is such that many word combinations are infrequent and do not occur in a given corpus. In this work we propose a method for estimating the probability of such previously unseen word combinations using available information on ``most similar'' words. We describe a probabilistic word association model based on distributional word similarity, and apply it to improving probability estimates for unseen word bigrams in a variant of Katz's back-off model. The similarity-based method yields a 20% perplexity improvement in the prediction of unseen bigrams and statistically significant reductions in speech-recognition error.
cmp-lg/9405002
Temporal Relations: Reference or Discourse Coherence?
cmp-lg cs.CL
The temporal relations that hold between events described by successive utterances are often left implicit or underspecified. We address the role of two phenomena with respect to the recovery of these relations: (1) the referential properties of tense, and (2) the role of temporal constraints imposed by coherence relations. We account for several facets of the identification of temporal relations through an integration of these.
cmp-lg/9405003
Some Bibliographical References on Intonation and Intonational Meaning
cmp-lg cs.CL
A by-no-means-complete collection of references for those interested in intonational meaning, with other miscellaneous references on intonation included. Additional references are welcome, and should be sent to julia@research.att.com.
cmp-lg/9405004
Syntactic-Head-Driven Generation
cmp-lg cs.CL
The previously proposed semantic-head-driven generation methods run into problems if none of the daughter constituents in the syntacto-semantic rule schemata of a grammar fits the definition of a semantic head given in Shieber et al. 1990. This is the case for the semantic analysis rules of certain constraint-based semantic representations, e.g. Underspecified Discourse Representation Structures (UDRSs) (Frank/Reyle 1992). Since head-driven generation in general has its merits, we simply return to a syntactic definition of `head' and demonstrate the feasibility of syntactic-head-driven generation. In addition to its generality, a syntactic-head-driven algorithm provides a basis for a logically well-defined treatment of the movement of (syntactic) heads, for which only ad-hoc solutions existed, so far.
cmp-lg/9405005
Pearl: A Probabilistic Chart Parser
cmp-lg cs.CL
This paper describes a natural language parsing algorithm for unrestricted text which uses a probability-based scoring function to select the "best" parse of a sentence. The parser, Pearl, is a time-asynchronous bottom-up chart parser with Earley-type top-down prediction which pursues the highest-scoring theory in the chart, where the score of a theory represents the extent to which the context of the sentence predicts that interpretation. This parser differs from previous attempts at stochastic parsers in that it uses a richer form of conditional probabilities based on context to predict likelihood. Pearl also provides a framework for incorporating the results of previous work in part-of-speech assignment, unknown word models, and other probabilistic models of linguistic features into one parsing tool, interleaving these techniques instead of using the traditional pipeline architecture. In preliminary tests, Pearl has been successful at resolving part-of-speech and word (in speech processing) ambiguity, determining categories for unknown words, and selecting correct parses first using a very loosely fitting covering grammar.
cmp-lg/9405006
Efficiency, Robustness, and Accuracy in Picky Chart Parsing
cmp-lg cs.CL
This paper describes Picky, a probabilistic agenda-based chart parsing algorithm which uses a technique called {\em probabilistic prediction} to predict which grammar rules are likely to lead to an acceptable parse of the input. Using a suboptimal search method, Picky significantly reduces the number of edges produced by CKY-like chart parsing algorithms, while maintaining the robustness of pure bottom-up parsers and the accuracy of existing probabilistic parsers. Experiments using Picky demonstrate how probabilistic modelling can impact upon the efficiency, robustness and accuracy of a parser.
cmp-lg/9405007
Towards History-based Grammars: Using Richer Models for Probabilistic Parsing
cmp-lg cs.CL
We describe a generative probabilistic model of natural language, which we call HBG, that takes advantage of detailed linguistic information to resolve ambiguity. HBG incorporates lexical, syntactic, semantic, and structural information from the parse tree into the disambiguation process in a novel way. We use a corpus of bracketed sentences, called a Treebank, in combination with decision tree building to tease out the relevant aspects of a parse tree that will determine the correct parse of a sentence. This stands in contrast to the usual approach of further grammar tailoring via the usual linguistic introspection in the hope of generating the correct parse. In head-to-head tests against one of the best existing robust probabilistic parsing models, which we call P-CFG, the HBG model significantly outperforms P-CFG, increasing the parsing accuracy rate from 60% to 75%, a 37% reduction in error.
cmp-lg/9405008
A Stochastic Finite-State Word-Segmentation Algorithm for Chinese
cmp-lg cs.CL
We present a stochastic finite-state model for segmenting Chinese text into dictionary entries and productively derived words, and providing pronunciations for these words; the method incorporates a class-based model in its treatment of personal names. We also evaluate the system's performance, taking into account the fact that people often do not agree on a single segmentation.
cmp-lg/9405009
Natural Language Parsing as Statistical Pattern Recognition
cmp-lg cs.CL
Traditional natural language parsers are based on rewrite rule systems developed in an arduous, time-consuming manner by grammarians. A majority of the grammarian's efforts are devoted to the disambiguation process, first hypothesizing rules which dictate constituent categories and relationships among words in ambiguous sentences, and then seeking exceptions and corrections to these rules. In this work, I propose an automatic method for acquiring a statistical parser from a set of parsed sentences which takes advantage of some initial linguistic input, but avoids the pitfalls of the iterative and seemingly endless grammar development process. Based on distributionally-derived and linguistically-based features of language, this parser acquires a set of statistical decision trees which assign a probability distribution on the space of parse trees given the input sentence. These decision trees take advantage of significant amount of contextual information, potentially including all of the lexical information in the sentence, to produce highly accurate statistical models of the disambiguation process. By basing the disambiguation criteria selection on entropy reduction rather than human intuition, this parser development method is able to consider more sentences than a human grammarian can when making individual disambiguation rules. In experiments between a parser, acquired using this statistical framework, and a grammarian's rule-based parser, developed over a ten-year period, both using the same training material and test sentences, the decision tree parser significantly outperformed the grammar-based parser on the accuracy measure which the grammarian was trying to maximize, achieving an accuracy of 78% compared to the grammar-based parser's 69%.
cmp-lg/9405010
Common Topics and Coherent Situations: Interpreting Ellipsis in the Context of Discourse Inference
cmp-lg cs.CL
It is claimed that a variety of facts concerning ellipsis, event reference, and interclausal coherence can be explained by two features of the linguistic form in question: (1) whether the form leaves behind an empty constituent in the syntax, and (2) whether the form is anaphoric in the semantics. It is proposed that these features interact with one of two types of discourse inference, namely {\it Common Topic} inference and {\it Coherent Situation} inference. The differing ways in which these types of inference utilize syntactic and semantic representations predicts phenomena for which it is otherwise difficult to account.
cmp-lg/9405011
A Plan-Based Model for Response Generation in Collaborative Task-Oriented Dialogues
cmp-lg cs.CL
This paper presents a plan-based architecture for response generation in collaborative consultation dialogues, with emphasis on cases in which the system (consultant) and user (executing agent) disagree. Our work contributes to an overall system for collaborative problem-solving by providing a plan-based framework that captures the {\em Propose-Evaluate-Modify} cycle of collaboration, and by allowing the system to initiate subdialogues to negotiate proposed additions to the shared plan and to provide support for its claims. In addition, our system handles in a unified manner the negotiation of proposed domain actions, proposed problem-solving actions, and beliefs proposed by discourse actions. Furthermore, it captures cooperative responses within the collaborative framework and accounts for why questions are sometimes never answered.
cmp-lg/9405012
Integration Of Visual Inter-word Constraints And Linguistic Knowledge In Degraded Text Recognition
cmp-lg cs.CL
Degraded text recognition is a difficult task. Given a noisy text image, a word recognizer can be applied to generate several candidates for each word image. High-level knowledge sources can then be used to select a decision from the candidate set for each word image. In this paper, we propose that visual inter-word constraints can be used to facilitate candidate selection. Visual inter-word constraints provide a way to link word images inside the text page, and to interpret them systematically.
cmp-lg/9405013
Collaboration on reference to objects that are not mutually known
cmp-lg cs.CL
In conversation, a person sometimes has to refer to an object that is not previously known to the other participant. We present a plan-based model of how agents collaborate on reference of this sort. In making a reference, an agent uses the most salient attributes of the referent. In understanding a reference, an agent determines his confidence in its adequacy as a means of identifying the referent. To collaborate, the agents use judgment, suggestion, and elaboration moves to refashion an inadequate referring expression.
cmp-lg/9405014
Classifying Cue Phrases in Text and Speech Using Machine Learning
cmp-lg cs.CL
Cue phrases may be used in a discourse sense to explicitly signal discourse structure, but also in a sentential sense to convey semantic rather than structural information. This paper explores the use of machine learning for classifying cue phrases as discourse or sentential. Two machine learning programs (Cgrendel and C4.5) are used to induce classification rules from sets of pre-classified cue phrases and their features. Machine learning is shown to be an effective technique for not only automating the generation of classification rules, but also for improving upon previous results.
cmp-lg/9405015
Intention-based Segmentation: Human Reliability and Correlation with Linguistic Cues
cmp-lg cs.CL
Certain spans of utterances in a discourse, referred to here as segments, are widely assumed to form coherent units. Further, the segmental structure of discourse has been claimed to constrain and be constrained by many phenomena. However, there is weak consensus on the nature of segments and the criteria for recognizing or generating them. We present quantitative results of a two part study using a corpus of spontaneous, narrative monologues. The first part evaluates the statistical reliability of human segmentation of our corpus, where speaker intention is the segmentation criterion. We then use the subjects' segmentations to evaluate the correlation of discourse segmentation with three linguistic cues (referential noun phrases, cue words, and pauses), using information retrieval metrics.
cmp-lg/9405016
Precise n-gram Probabilities from Stochastic Context-free Grammars
cmp-lg cs.CL
We present an algorithm for computing n-gram probabilities from stochastic context-free grammars, a procedure that can alleviate some of the standard problems associated with n-grams (estimation from sparse data, lack of linguistic structure, among others). The method operates via the computation of substring expectations, which in turn is accomplished by solving systems of linear equations derived from the grammar. We discuss efficient implementation of the algorithm and report our practical experience with it.
cmp-lg/9405017
Best-first Model Merging for Hidden Markov Model Induction
cmp-lg cs.CL
This report describes a new technique for inducing the structure of Hidden Markov Models from data which is based on the general `model merging' strategy (Omohundro 1992). The process begins with a maximum likelihood HMM that directly encodes the training data. Successively more general models are produced by merging HMM states. A Bayesian posterior probability criterion is used to determine which states to merge and when to stop generalizing. The procedure may be considered a heuristic search for the HMM structure with the highest posterior probability. We discuss a variety of possible priors for HMMs, as well as a number of approximations which improve the computational efficiency of the algorithm. We studied three applications to evaluate the procedure. The first compares the merging algorithm with the standard Baum-Welch approach in inducing simple finite-state languages from small, positive-only training samples. We found that the merging procedure is more robust and accurate, particularly with a small amount of training data. The second application uses labelled speech data from the TIMIT database to build compact, multiple-pronunciation word models that can be used in speech recognition. Finally, we describe how the algorithm was incorporated in an operational speech understanding system, where it is combined with neural network acoustic likelihood estimators to improve performance over single-pronunciation word models.
cmp-lg/9405018
Memory-Based Lexical Acquisition and Processing
cmp-lg cs.CL
Current approaches to computational lexicology in language technology are knowledge-based (competence-oriented) and try to abstract away from specific formalisms, domains, and applications. This results in severe complexity, acquisition and reusability bottlenecks. As an alternative, we propose a particular performance-oriented approach to Natural Language Processing based on automatic memory-based learning of linguistic (lexical) tasks. The consequences of the approach for computational lexicology are discussed, and the application of the approach on a number of lexical acquisition and disambiguation tasks in phonology, morphology and syntax is described.
cmp-lg/9405019
Determination of referential property and number of nouns in Japanese sentences for machine translation into English
cmp-lg cs.CL
When translating Japanese nouns into English, we face the problem of articles and numbers which the Japanese language does not have, but which are necessary for the English composition. To solve this difficult problem we classified the referential property and the number of nouns into three types respectively. This paper shows that the referential property and the number of nouns in a sentence can be estimated fairly reliably by the words in the sentence. Many rules for the estimation were written in forms similar to rewriting rules in expert systems. We obtained the correct recognition scores of 85.5\% and 89.0\% in the estimation of the referential property and the number respectively for the sentences which were used for the construction of our rules. We tested these rules for some other texts, and obtained the scores of 68.9\% and 85.6\% respectively.
cmp-lg/9405020
Capturing CFLs with Tree Adjoining Grammars
cmp-lg cs.CL
We define a decidable class of TAGs that is strongly equivalent to CFGs and is cubic-time parsable. This class serves to lexicalize CFGs in the same manner as the LCFGs of Schabes and Waters but with considerably less restriction on the form of the grammars. The class provides a normal form for TAGs that generate local sets in much the same way that regular grammars provide a normal form for CFGs that generate regular sets.
cmp-lg/9405021
Generating Precondition Expressions in Instructional Text
cmp-lg cs.CL
This study employs a knowledge intensive corpus analysis to identify the elements of the communicative context which can be used to determine the appropriate lexical and grammatical form of instructional texts. \ig, an instructional text generation system based on this analysis, is presented, particularly with reference to its expression of precondition relations.
cmp-lg/9405022
Grammar Specialization through Entropy Thresholds
cmp-lg cs.CL
Explanation-based generalization is used to extract a specialized grammar from the original one using a training corpus of parse trees. This allows very much faster parsing and gives a lower error rate, at the price of a small loss in coverage. Previously, it has been necessary to specify the tree-cutting criteria (or operationality criteria) manually; here they are derived automatically from the training set and the desired coverage of the specialized grammar. This is done by assigning an entropy value to each node in the parse trees and cutting in the nodes with sufficiently high entropy values.
cmp-lg/9405023
An Integrated Heuristic Scheme for Partial Parse Evaluation
cmp-lg cs.CL
GLR* is a recently developed robust version of the Generalized LR Parser, that can parse almost ANY input sentence by ignoring unrecognizable parts of the sentence. On a given input sentence, the parser returns a collection of parses that correspond to maximal, or close to maximal, parsable subsets of the original input. This paper describes recent work on developing an integrated heuristic scheme for selecting the parse that is deemed ``best'' from such a collection. We describe the heuristic measures used and their combination scheme. Preliminary results from experiments conducted on parsing speech recognized spontaneous speech are also reported.
cmp-lg/9405024
Abductive Equivalential Translation and its application to Natural Language Database Interfacing
cmp-lg cs.CL
The thesis describes a logical formalization of natural-language database interfacing. We assume the existence of a ``natural language engine'' capable of mediating between surface linguistic string and their representations as ``literal'' logical forms: the focus of interest will be the question of relating ``literal'' logical forms to representations in terms of primitives meaningful to the underlying database engine. We begin by describing the nature of the problem, and show how a variety of interface functionalities can be considered as instances of a type of formal inference task which we call ``Abductive Equivalential Translation'' (AET); functionalities which can be reduced to this form include answering questions, responding to commands, reasoning about the completeness of answers, answering meta-questions of type ``Do you know...'', and generating assertions and questions. In each case, a ``linguistic domain theory'' (LDT) $\Gamma$ and an input formula $F$ are given, and the goal is to construct a formula with certain properties which is equivalent to $F$, given $\Gamma$ and a set of permitted assumptions. If the LDT is of a certain specified type, whose formulas are either conditional equivalences or Horn-clauses, we show that the AET problem can be reduced to a goal-directed inference method. We present an abstract description of this method, and sketch its realization in Prolog. The relationship between AET and several problems previously discussed in the literature is discussed. In particular, we show how AET can provide a simple and elegant solution to the so-called ``Doctor on Board'' problem, and in effect allows a ``relativization'' of the Closed World Assumption. The ideas in the thesis have all been implemented concretely within the SRI CLARE project, using a real projects and payments database. The LDT for the example database is described in detail, and examples of the types of functionality that can be achieved within the example domain are presented.
cmp-lg/9405025
An Optimal Tabular Parsing Algorithm
cmp-lg cs.CL
In this paper we relate a number of parsing algorithms which have been developed in very different areas of parsing theory, and which include deterministic algorithms, tabular algorithms, and a parallel algorithm. We show that these algorithms are based on the same underlying ideas. By relating existing ideas, we hope to provide an opportunity to improve some algorithms based on features of others. A second purpose of this paper is to answer a question which has come up in the area of tabular parsing, namely how to obtain a parsing algorithm with the property that the table will contain as little entries as possible, but without the possibility that two entries represent the same subderivation.
cmp-lg/9405026
An Extended Theory of Head-Driven Parsing
cmp-lg cs.CL
We show that more head-driven parsing algorithms can be formulated than those occurring in the existing literature. These algorithms are inspired by a family of left-to-right parsing algorithms from a recent publication. We further introduce a more advanced notion of ``head-driven parsing'' which allows more detailed specification of the processing order of non-head elements in the right-hand side. We develop a parsing algorithm for this strategy, based on LR parsing techniques.
cmp-lg/9405027
Acquiring Receptive Morphology: A Connectionist Model
cmp-lg cs.CL
This paper describes a modular connectionist model of the acquisition of receptive inflectional morphology. The model takes inputs in the form of phones one at a time and outputs the associated roots and inflections. Simulations using artificial language stimuli demonstrate the capacity of the model to learn suffixation, prefixation, infixation, circumfixation, mutation, template, and deletion rules. Separate network modules responsible for syllables enable to the network to learn simple reduplication rules as well. The model also embodies constraints against association-line crossing.
cmp-lg/9405028
Semantics of Complex Sentences in Japanese
cmp-lg cs.CL
The important part of semantics of complex sentence is captured as relations among semantic roles in subordinate and main clause respectively. However if there can be relations between every pair of semantic roles, the amount of computation to identify the relations that hold in the given sentence is extremely large. In this paper, for semantics of Japanese complex sentence, we introduce new pragmatic roles called `observer' and `motivated' respectively to bridge semantic roles of subordinate and those of main clauses. By these new roles constraints on the relations among semantic/pragmatic roles are known to be almost local within subordinate or main clause. In other words, as for the semantics of the whole complex sentence, the only role we should deal with is a motivated.
cmp-lg/9405029
Structural Tags, Annealing and Automatic Word Classification
cmp-lg cs.CL
This paper describes an automatic word classification system which uses a locally optimal annealing algorithm and average class mutual information. A new word-class representation, the structural tag is introduced and its advantages for use in statistical language modelling are presented. A summary of some results with the one million word LOB corpus is given; the algorithm is also shown to discover the vowel-consonant distinction and displays an ability to cluster words syntactically in a Latin corpus. Finally, a comparison is made between the current classification system and several leading alternative systems, which shows that the current system performs respectably well.
cmp-lg/9405030
Priority Union and Generalization in Discourse Grammars
cmp-lg cs.CL
We describe an implementation in Carpenter's typed feature formalism, ALE, of a discourse grammar of the kind proposed by Scha, Polanyi, et al. We examine their method for resolving parallelism-dependent anaphora and show that there is a coherent feature-structural rendition of this type of grammar which uses the operations of priority union and generalization. We describe an augmentation of the ALE system to encompass these operations and we show that an appropriate choice of definition for priority union gives the desired multiple output for examples of VP-ellipsis which exhibit a strict/sloppy ambiguity.
cmp-lg/9405031
An Attributive Logic of Set Descriptions and Set Operations
cmp-lg cs.CL
This paper provides a model theoretic semantics to feature terms augmented with set descriptions. We provide constraints to specify HPSG style set descriptions, fixed cardinality set descriptions, set-membership constraints, restricted universal role quantifications, set union, intersection, subset and disjointness. A sound, complete and terminating consistency checking procedure is provided to determine the consistency of any given term in the logic. It is shown that determining consistency of terms is a NP-complete problem.
cmp-lg/9405032
Modularity in a Connectionist Model of Morphology Acquisition
cmp-lg cs.CL
This paper describes a modular connectionist model of the acquisition of receptive inflectional morphology. The model takes inputs in the form of phones one at a time and outputs the associated roots and inflections. In its simplest version, the network consists of separate simple recurrent subnetworks for root and inflection identification; both networks take the phone sequence as inputs. It is shown that the performance of the two separate modular networks is superior to a single network responsible for both root and inflection identification. In a more elaborate version of the model, the network learns to use separate hidden-layer modules to solve the separate tasks of root and inflection identification.
cmp-lg/9405033
Relating Complexity to Practical Performance in Parsing with Wide-Coverage Unification Grammars
cmp-lg cs.CL
The paper demonstrates that exponential complexities with respect to grammar size and input length have little impact on the performance of three unification-based parsing algorithms, using a wide-coverage grammar. The results imply that the study and optimisation of unification-based parsing must rely on empirical data until complexity theory can more accurately predict the practical behaviour of such parsers.
cmp-lg/9405034
Extracting Noun Phrases from Large-Scale Texts: A Hybrid Approach and Its Automatic Evaluation
cmp-lg cs.CL
To acquire noun phrases from running texts is useful for many applications, such as word grouping,terminology indexing, etc. The reported literatures adopt pure probabilistic approach, or pure rule-based noun phrases grammar to tackle this problem. In this paper, we apply a probabilistic chunker to deciding the implicit boundaries of constituents and utilize the linguistic knowledge to extract the noun phrases by a finite state mechanism. The test texts are SUSANNE Corpus and the results are evaluated by comparing the parse field of SUSANNE Corpus automatically. The results of this preliminary experiment are encouraging.
cmp-lg/9405035
Dual-Coding Theory and Connectionist Lexical Selection
cmp-lg cs.CL
We introduce the bilingual dual-coding theory as a model for bilingual mental representation. Based on this model, lexical selection neural networks are implemented for a connectionist transfer project in machine translation. This lexical selection approach has two advantages. First, it is learnable. Little human effort on knowledge engineering is required. Secondly, it is psycholinguistically well-founded.
cmp-lg/9406001
Intentions and Information in Discourse
cmp-lg cs.CL
This paper is about the flow of inference between communicative intentions, discourse structure and the domain during discourse processing. We augment a theory of discourse interpretation with a theory of distinct mental attitudes and reasoning about them, in order to provide an account of how the attitudes interact with reasoning about discourse structure.
cmp-lg/9406002
Speech Dialogue with Facial Displays: Multimodal Human-Computer Conversation
cmp-lg cs.CL
Human face-to-face conversation is an ideal model for human-computer dialogue. One of the major features of face-to-face communication is its multiplicity of communication channels that act on multiple modalities. To realize a natural multimodal dialogue, it is necessary to study how humans perceive information and determine the information to which humans are sensitive. A face is an independent communication channel that conveys emotional and conversational signals, encoded as facial expressions. We have developed an experimental system that integrates speech dialogue and facial animation, to investigate the effect of introducing communicative facial expressions as a new modality in human-computer conversation. Our experiments have shown that facial expressions are helpful, especially upon first contact with the system. We have also discovered that featuring facial expressions at an early stage improves subsequent interaction.
cmp-lg/9406003
A Learning Approach to Natural Language Understanding
cmp-lg cs.CL
In this paper we propose a learning paradigm for the problem of understanding spoken language. The basis of the work is in a formalization of the understanding problem as a communication problem. This results in the definition of a stochastic model of the production of speech or text starting from the meaning of a sentence. The resulting understanding algorithm consists in a Viterbi maximization procedure, analogous to that commonly used for recognizing speech. The algorithm was implemented for building
cmp-lg/9406004
Towards a Principled Representation of Discourse Plans
cmp-lg cs.CL
We argue that discourse plans must capture the intended causal and decompositional relations between communicative actions. We present a planning algorithm, DPOCL, that builds plan structures that properly capture these relations, and show how these structures are used to solve the problems that plagued previous discourse planners, and allow a system to participate effectively and flexibly in an ongoing dialogue.
cmp-lg/9406005
Word-Sense Disambiguation Using Decomposable Models
cmp-lg cs.CL
Most probabilistic classifiers used for word-sense disambiguation have either been based on only one contextual feature or have used a model that is simply assumed to characterize the interdependencies among multiple contextual features. In this paper, a different approach to formulating a probabilistic model is presented along with a case study of the performance of models produced in this manner for the disambiguation of the noun "interest". We describe a method for formulating probabilistic models that use multiple contextual features for word-sense disambiguation, without requiring untested assumptions regarding the form of the model. Using this approach, the joint distribution of all variables is described by only the most systematic variable interactions, thereby limiting the number of parameters to be estimated, supporting computational efficiency, and providing an understanding of the data.
cmp-lg/9406006
Detecting and Correcting Speech Repairs
cmp-lg cs.CL
Interactive spoken dialog provides many new challenges for spoken language systems. One of the most critical is the prevalence of speech repairs. This paper presents an algorithm that detects and corrects speech repairs based on finding the repair pattern. The repair pattern is built by finding word matches and word replacements, and identifying fragments and editing terms. Rather than using a set of prebuilt templates, we build the pattern on the fly. In a fair test, our method, when combined with a statistical model to filter possible repairs, was successful at detecting and correcting 80\% of the repairs, without using prosodic information or a parser.
cmp-lg/9406007
Aligning a Parallel English-Chinese Corpus Statistically with Lexical Criteria
cmp-lg cs.CL
We describe our experience with automatic alignment of sentences in parallel English-Chinese texts. Our report concerns three related topics: (1) progress on the HKUST English-Chinese Parallel Bilingual Corpus; (2) experiments addressing the applicability of Gale & Church's length-based statistical method to the task of alignment involving a non-Indo-European language; and (3) an improved statistical method that also incorporates domain-specific lexical cues.
cmp-lg/9406008
Parsing Turkish with the Lexical Functional Grammar Formalism
cmp-lg cs.CL
This paper describes our work on parsing Turkish using the lexical-functional grammar formalism. This work represents the first significant effort for parsing Turkish. Our implementation is based on Tomita's parser developed at Carnegie-Mellon University Center for Machine Translation. The grammar covers a substantial subset of Turkish including simple and complex sentences, and deals with a reasonable amount of word order freeness. The complex agglutinative morphology of Turkish lexical structures is handled using a separate two-level morphological analyzer. After a discussion of key relevant issues regarding Turkish grammar, we discuss aspects of our system and present results from our implementation. Our initial results suggest that our system can parse about 82\% of the sentences directly and almost all the remaining with very minor pre-editing.
cmp-lg/9406009
Multiset-Valued Linear Index Grammars: Imposing Dominance Constraints on Derivations
cmp-lg cs.CL
This paper defines multiset-valued linear index grammar and unordered vector grammar with dominance links. The former models certain uses of multiset-valued feature structures in unification-based formalisms, while the latter is motivated by word order variation and by ``quasi-trees'', a generalization of trees. The two formalisms are weakly equivalent, and an important subset is at most context-sensitive and polynomially parsable.
cmp-lg/9406010
Some Advances in Transformation-Based Part of Speech Tagging
cmp-lg cs.CL
Most recent research in trainable part of speech taggers has explored stochastic tagging. While these taggers obtain high accuracy, linguistic information is captured indirectly, typically in tens of thousands of lexical and contextual probabilities. In [Brill92], a trainable rule-based tagger was described that obtained performance comparable to that of stochastic taggers, but captured relevant linguistic information in a small number of simple non-stochastic rules. In this paper, we describe a number of extensions to this rule-based tagger. First, we describe a method for expressing lexical relations in tagging that are not captured by stochastic taggers. Next, we show a rule-based approach to tagging unknown words. Finally, we show how the tagger can be extended into a k-best tagger, where multiple tags can be assigned to words in some cases of uncertainty.
cmp-lg/9406011
Exploring the Statistical Derivation of Transformational Rule Sequences for Part-of-Speech Tagging
cmp-lg cs.CL
Eric Brill has recently proposed a simple and powerful corpus-based language modeling approach that can be applied to various tasks including part-of-speech tagging and building phrase structure trees. The method learns a series of symbolic transformational rules, which can then be applied in sequence to a test corpus to produce predictions. The learning process only requires counting matches for a given set of rule templates, allowing the method to survey a very large space of possible contextual factors. This paper analyses Brill's approach as an interesting variation on existing decision tree methods, based on experiments involving part-of-speech tagging for both English and ancient Greek corpora. In particular, the analysis throws light on why the new mechanism seems surprisingly resistant to overtraining. A fast, incremental implementation and a mechanism for recording the dependencies that underlie the resulting rule sequence are also described.
cmp-lg/9406012
Self-Organizing Machine Translation: Example-Driven Induction of Transfer Functions
cmp-lg cs.CL
With the advent of faster computers, the notion of doing machine translation from a huge stored database of translation examples is no longer unreasonable. This paper describes an attempt to merge the Example-Based Machine Translation (EBMT) approach with psycholinguistic principles. A new formalism for context- free grammars, called *marker-normal form*, is demonstrated and used to describe language data in a way compatible with psycholinguistic theories. By embedding this formalism in a standard multivariate optimization framework, a system can be built that infers correct transfer functions for a set of bilingual sentence pairs and then uses those functions to translate novel sentences. The validity of this line of reasoning has been tested in the development of a system called METLA-1. This system has been used to infer English->French and English->Urdu transfer functions from small corpora. The results of those experiments are examined, both in engineering terms as well as in more linguistic terms. In general, the results of these experiments were psycho- logically and linguistically well-grounded while still achieving a respectable level of success when compared against a similar prototype using Hidden Markov Models.
cmp-lg/9406013
Graded Unification: A Framework for Interactive Processing
cmp-lg cs.CL
An extension to classical unification, called {\em graded unification} is presented. It is capable of combining contradictory information. An interactive processing paradigm and parser based on this new operator are also presented.
cmp-lg/9406014
A Hybrid Reasoning Model for Indirect Answers
cmp-lg cs.CL
This paper presents our implemented computational model for interpreting and generating indirect answers to Yes-No questions. Its main features are 1) a discourse-plan-based approach to implicature, 2) a reversible architecture for generation and interpretation, 3) a hybrid reasoning model that employs both plan inference and logical inference, and 4) use of stimulus conditions to model a speaker's motivation for providing appropriate, unrequested information. The model handles a wider range of types of indirect answers than previous computational models and has several significant advantages.
cmp-lg/9406015
Statistical Augmentation of a Chinese Machine-Readable Dictionary
cmp-lg cs.CL
We describe a method of using statistically-collected Chinese character groups from a corpus to augment a Chinese dictionary. The method is particularly useful for extracting domain-specific and regional words not readily available in machine-readable dictionaries. Output was evaluated both using human evaluators and against a previously available dictionary. We also evaluated performance improvement in automatic Chinese tokenization. Results show that our method outputs legitimate words, acronymic constructions, idioms, names and titles, as well as technical compounds, many of which were lacking from the original dictionary.
cmp-lg/9406016
Corpus-Driven Knowledge Acquisition for Discourse Analysis
cmp-lg cs.CL
The availability of large on-line text corpora provides a natural and promising bridge between the worlds of natural language processing (NLP) and machine learning (ML). In recent years, the NLP community has been aggressively investigating statistical techniques to drive part-of-speech taggers, but application-specific text corpora can be used to drive knowledge acquisition at much higher levels as well. In this paper we will show how ML techniques can be used to support knowledge acquisition for information extraction systems. It is often very difficult to specify an explicit domain model for many information extraction applications, and it is always labor intensive to implement hand-coded heuristics for each new domain. We have discovered that it is nevertheless possible to use ML algorithms in order to capture knowledge that is only implicitly present in a representative text corpus. Our work addresses issues traditionally associated with discourse analysis and intersentential inference generation, and demonstrates the utility of ML algorithms at this higher level of language analysis. The benefits of our work address the portability and scalability of information extraction (IE) technologies. When hand-coded heuristics are used to manage discourse analysis in an information extraction system, months of programming effort are easily needed to port a successful IE system to a new domain. We will show how ML algorithms can reduce this
cmp-lg/9406017
An Automatic Method of Finding Topic Boundaries
cmp-lg cs.CL
This article outlines a new method of locating discourse boundaries based on lexical cohesion and a graphical technique called dotplotting. The application of dotplotting to discourse segmentation can be performed either manually, by examining a graph, or automatically, using an optimization algorithm. The results of two experiments involving automatically locating boundaries between a series of concatenated documents are presented. Areas of application and future directions for this work are also outlined.
cmp-lg/9406018
TDL--- A Type Description Language for Constraint-Based Grammars
cmp-lg cs.CL
This paper presents \tdl, a typed feature-based representation language and inference system. Type definitions in \tdl\ consist of type and feature constraints over the boolean connectives. \tdl\ supports open- and closed-world reasoning over types and allows for partitions and incompatible types. Working with partially as well as with fully expanded types is possible. Efficient reasoning in \tdl\ is accomplished through specialized modules.
cmp-lg/9406019
A Complete and Recursive Feature Theory
cmp-lg cs.CL
Various feature descriptions are being employed in logic programming languages and constrained-based grammar formalisms. The common notational primitive of these descriptions are functional attributes called features. The descriptions considered in this paper are the possibly quantified first-order formulae obtained from a signature of binary and unary predicates called features and sorts, respectively. We establish a first-order theory FT by means of three axiom schemes, show its completeness, and construct three elementarily equivalent models. One of the models consists of so-called feature graphs, a data structure common in computational linguistics. The other two models consist of so-called feature trees, a record-like data structure generalizing the trees corresponding to first-order terms. Our completeness proof exhibits a terminating simplification system deciding validity and satisfiability of possibly quantified feature descriptions.
cmp-lg/9406020
DPOCL: A Principled Approach to Discourse Planning
cmp-lg cs.CL
Research in discourse processing has identified two representational requirements for discourse planning systems. First, discourse plans must adequately represent the intentional structure of the utterances they produce in order to enable a computational discourse agent to respond effectively to communicative failures \cite{MooreParisCL}. Second, discourse plans must represent the informational structure of utterances. In addition to these representational requirements, we argue that discourse planners should be formally characterizable in terms of soundness and completeness.