{ "paper_id": "P06-1002", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T09:24:23.971890Z" }, "title": "Going Beyond AER: An Extensive Analysis of Word Alignments and Their Impact on MT", "authors": [ { "first": "Necip", "middle": [], "last": "Fazil", "suffix": "", "affiliation": { "laboratory": "", "institution": "University of Maryland College Park", "location": { "postCode": "20742", "region": "MD" } }, "email": "" }, { "first": "Bonnie", "middle": [ "J" ], "last": "Dorr", "suffix": "", "affiliation": { "laboratory": "", "institution": "University of Maryland College Park", "location": { "postCode": "20742", "region": "MD" } }, "email": "bonnie@umiacs.umd.edu" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "This paper presents an extensive evaluation of five different alignments and investigates their impact on the corresponding MT system output. We introduce new measures for intrinsic evaluations and examine the distribution of phrases and untranslated words during decoding to identify which characteristics of different alignments affect translation. We show that precision-oriented alignments yield better MT output (translating more words and using longer phrases) than recalloriented alignments.", "pdf_parse": { "paper_id": "P06-1002", "_pdf_hash": "", "abstract": [ { "text": "This paper presents an extensive evaluation of five different alignments and investigates their impact on the corresponding MT system output. We introduce new measures for intrinsic evaluations and examine the distribution of phrases and untranslated words during decoding to identify which characteristics of different alignments affect translation. We show that precision-oriented alignments yield better MT output (translating more words and using longer phrases) than recalloriented alignments.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "Word alignments are a by-product of statistical machine translation (MT) and play a crucial role in MT performance. In recent years, researchers have proposed several algorithms to generate word alignments. However, evaluating word alignments is difficult because even humans have difficulty performing this task.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "The state-of-the art evaluation metricalignment error rate (AER)-attempts to balance the precision and recall scores at the level of alignment links (Och and Ney, 2000) . Other metrics assess the impact of alignments externally, e.g., different alignments are tested by comparing the corresponding MT outputs using automated evaluation metrics (e.g., BLEU (Papineni et al., 2002) or METEOR (Banerjee and Lavie, 2005) ). However, these studies showed that AER and BLEU do not correlate well (Callison-Burch et al., 2004; Goutte et al., 2004; Ittycheriah and Roukos, 2005) . Despite significant AER improvements achieved by several researchers, the improvements in BLEU scores are insignificant or, at best, small. This paper demonstrates the difficulty in assessing whether alignment quality makes a difference in MT performance. We describe the impact of certain alignment characteristics on MT performance but also identify several alignment-related factors that impact MT performance regardless of the quality of the initial alignments. In so doing, we begin to answer long-standing questions about the value of alignment in the context of MT.", "cite_spans": [ { "start": 149, "end": 168, "text": "(Och and Ney, 2000)", "ref_id": "BIBREF12" }, { "start": 356, "end": 379, "text": "(Papineni et al., 2002)", "ref_id": "BIBREF16" }, { "start": 390, "end": 416, "text": "(Banerjee and Lavie, 2005)", "ref_id": "BIBREF1" }, { "start": 490, "end": 519, "text": "(Callison-Burch et al., 2004;", "ref_id": "BIBREF3" }, { "start": 520, "end": 540, "text": "Goutte et al., 2004;", "ref_id": "BIBREF5" }, { "start": 541, "end": 570, "text": "Ittycheriah and Roukos, 2005)", "ref_id": "BIBREF6" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "We first evaluate 5 different word alignments intrinsically, using: (1) community-standard metrics-precision, recall and AER; and (2) a new measure called consistent phrase error rate (CPER). Next, we observe the impact of different alignments on MT performance. We present BLEU scores on a phrase-based MT system, Pharaoh (Koehn, 2004) , using five different alignments to extract phrases. We investigate the impact of different settings for phrase extraction, lexical weighting, maximum phrase length and training data. Finally, we present a quantitative analysis of which phrases are chosen during the actual decoding process and show how the distribution of the phrases differ from one alignment into another.", "cite_spans": [ { "start": 323, "end": 336, "text": "(Koehn, 2004)", "ref_id": "BIBREF8" } ], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Our experiments show that precision-oriented alignments yield better phrases for MT than recalloriented alignments. Specifically, they cover a higher percentage of our test sets and result in fewer untranslated words and selection of longer phrases during decoding.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "The next section describes work related to our alignment evaluation approach. Following this we outline different intrinsic evaluation measures of alignment and we propose a new measure to evaluate word alignments within phrase-based MT framework. We then present several experiments to measure the impact of different word alignments on a phrase-based MT system, and investigate how different alignments change the phrase selection in the same MT system.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Introduction", "sec_num": "1" }, { "text": "Starting with the IBM models (Brown et al., 1993) , researchers have developed various statistical word alignment systems based on different models, such as hidden Markov models (HMM) (Vogel et al., 1996) , log-linear models (Och and Ney, 2003) , and similarity-based heuristic methods (Melamed, 2000) . These methods are unsupervised, i.e., the only input is large parallel corpora. In recent years, researchers have shown that even using a limited amount of manually aligned data improves word alignment significantly (Callison-Burch et al., 2004) . Supervised learning techniques, such as perceptron learning, maximum entropy modeling or maximum weighted bipartite matching, have been shown to provide further improvements on word alignments (Ayan et al., 2005; Moore, 2005; Ittycheriah and Roukos, 2005; Taskar et al., 2005) .", "cite_spans": [ { "start": 29, "end": 49, "text": "(Brown et al., 1993)", "ref_id": "BIBREF2" }, { "start": 184, "end": 204, "text": "(Vogel et al., 1996)", "ref_id": "BIBREF18" }, { "start": 225, "end": 244, "text": "(Och and Ney, 2003)", "ref_id": "BIBREF14" }, { "start": 286, "end": 301, "text": "(Melamed, 2000)", "ref_id": "BIBREF10" }, { "start": 520, "end": 549, "text": "(Callison-Burch et al., 2004)", "ref_id": "BIBREF3" }, { "start": 745, "end": 764, "text": "(Ayan et al., 2005;", "ref_id": "BIBREF0" }, { "start": 765, "end": 777, "text": "Moore, 2005;", "ref_id": "BIBREF11" }, { "start": 778, "end": 807, "text": "Ittycheriah and Roukos, 2005;", "ref_id": "BIBREF6" }, { "start": 808, "end": 828, "text": "Taskar et al., 2005)", "ref_id": "BIBREF17" } ], "ref_spans": [], "eq_spans": [], "section": "Related Work", "sec_num": "2" }, { "text": "The standard technique for evaluating word alignments is to represent alignments as a set of links (i.e., pairs of words) and to compare the generated alignment against manual alignment of the same data at the level of links. Manual alignments are represented by two sets: Probable (P ) alignments and Sure (S) alignments, where S \u2286 P . Given A, P and S, the most commonly used metrics-precision (Pr), recall (Rc) and alignment error rate (AER)-are defined as follows:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Related Work", "sec_num": "2" }, { "text": "P r = |A \u2229 P | |A| Rc = |A \u2229 S| |S| AER = 1 \u2212 |A \u2229 S| + |A \u2229 P | |A| + |S|", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Related Work", "sec_num": "2" }, { "text": "Another approach to evaluating alignments is to measure their impact on an external application, e.g., statistical MT. In recent years, phrase-based systems (Koehn, 2004; Chiang, 2005) have been shown to outperform word-based MT systems; therefore, in this paper, we use a publicly-available phrase-based MT system, Pharaoh (Koehn, 2004) , to investigate the impact of different alignments.", "cite_spans": [ { "start": 157, "end": 170, "text": "(Koehn, 2004;", "ref_id": "BIBREF8" }, { "start": 171, "end": 184, "text": "Chiang, 2005)", "ref_id": "BIBREF4" }, { "start": 324, "end": 337, "text": "(Koehn, 2004)", "ref_id": "BIBREF8" } ], "ref_spans": [], "eq_spans": [], "section": "Related Work", "sec_num": "2" }, { "text": "Although it is possible to estimate phrases directly from a training corpus (Marcu and Wong, 2002) , most phrase-based MT systems (Koehn, 2004; Chiang, 2005) start with a word alignment and extract phrases that are consistent with the given alignment. Once the consistent phrases are extracted, they are assigned multiple scores (such as translation probabilities and lexical weights), and the decoder's job is to choose the correct phrases based on those scores using a log-linear model.", "cite_spans": [ { "start": 76, "end": 98, "text": "(Marcu and Wong, 2002)", "ref_id": "BIBREF9" }, { "start": 130, "end": 143, "text": "(Koehn, 2004;", "ref_id": "BIBREF8" }, { "start": 144, "end": 157, "text": "Chiang, 2005)", "ref_id": "BIBREF4" } ], "ref_spans": [], "eq_spans": [], "section": "Related Work", "sec_num": "2" }, { "text": "Our goal is to compare different alignments and to investigate how their characteristics affect the MT systems. We evaluate alignments in terms of precision, recall, alignment error rate (AER), and a new measure called consistent phrase error rate (CPER). We focus on 5 different alignments obtained by combining two uni-directional alignments. Each uni-directional alignment is the result of running GIZA++ (Och, 2000b) in one of two directions (source-to-target and vice versa) with default configurations. The combined alignments that are used in this paper are as follows:", "cite_spans": [ { "start": 408, "end": 420, "text": "(Och, 2000b)", "ref_id": "BIBREF13" } ], "ref_spans": [], "eq_spans": [], "section": "Intrinsic Evaluation of Alignments", "sec_num": "3" }, { "text": "1. Union of both directions (S U ), 2. Intersection of both directions (S I ), 3. A heuristic based combination technique called grow-diag-final (S G ), which is the default alignment combination heuristic employed in Pharaoh (Koehn, 2004) , 4-5. Two supervised alignment combination techniques (S A and S B ) using 2 and 4 input alignments as described in (Ayan et al., 2005) . This paper examines the impact of alignments according to their orientation toward precision or recall. Among the five alignments above, S U and S G are recall-oriented while the other three are precision-oriented. S B is an improved version of S A which attempts to increase recall without a significant sacrifice in precision.", "cite_spans": [ { "start": 226, "end": 239, "text": "(Koehn, 2004)", "ref_id": "BIBREF8" }, { "start": 357, "end": 376, "text": "(Ayan et al., 2005)", "ref_id": "BIBREF0" } ], "ref_spans": [], "eq_spans": [], "section": "Intrinsic Evaluation of Alignments", "sec_num": "3" }, { "text": "Manually aligned data from two language pairs are used in our intrinsic evaluations using the five combinations above. A summary of the training and test data is presented in Table 1 .", "cite_spans": [], "ref_spans": [ { "start": 175, "end": 182, "text": "Table 1", "ref_id": "TABREF1" } ], "eq_spans": [], "section": "Intrinsic Evaluation of Alignments", "sec_num": "3" }, { "text": "Our gold standard for each language pair is a manually aligned corpus. English-Chinese an-notations distinguish between sure and probable alignment links, but English-Arabic annotations do not. The details of how the annotations are done can be found in (Ayan et al., 2005) and (Ittycheriah and Roukos, 2005) . Table 2 presents the precision, recall, and AER for 5 different alignments on 2 language pairs. For each of these metrics, a different system achieves the best score -respectively, these are S I , S U , and S B . S U and S G yield low precision, high recall alignments. In contrast, S I yields very high precision but very low recall. S A and S B attempt to balance these two measures but their precision is still higher than their recall. Both systems have nearly the same precision but S B yields significantly higher recall than S A .", "cite_spans": [ { "start": 254, "end": 273, "text": "(Ayan et al., 2005)", "ref_id": "BIBREF0" }, { "start": 278, "end": 308, "text": "(Ittycheriah and Roukos, 2005)", "ref_id": "BIBREF6" } ], "ref_spans": [ { "start": 311, "end": 318, "text": "Table 2", "ref_id": "TABREF3" } ], "eq_spans": [], "section": "Intrinsic Evaluation of Alignments", "sec_num": "3" }, { "text": "Align.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Precision, Recall and AER", "sec_num": "3.1" }, { "text": "en-ch en-ar Sys.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Precision, Recall and AER", "sec_num": "3.1" }, { "text": "Pr ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Precision, Recall and AER", "sec_num": "3.1" }, { "text": "In this section, we present a new method, called consistent phrase error rate (CPER), for evaluating word alignments in the context of phrasebased MT. The idea is to compare phrases consistent with a given alignment against phrases that would be consistent with human alignments. CPER is similar to AER but operates at the phrase level instead of at the word level. To compute CPER, we define a link in terms of the position of its start and end words in the phrases. For instance, the phrase link (i 1 , i 2 , j 1 , j 2 ) indicates that the English phrase e i 1 , . . . , e i 2 and the FL phrase f j 1 , . . . , f j 2 are consistent with the given alignment. Once we generate the set of phrases P A and P G that are consistent with a given alignment A and a manual alignment G, respectively, we compute precision (Pr), recall (Rc), and CPER as follows: 1", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Consistent Phrase Error Rate", "sec_num": "3.2" }, { "text": "P r = |P A \u2229 P G | |P A | Rc = |P A \u2229 P G | |P G | CP ER = 1 \u2212 2 \u00d7 P r \u00d7 Rc P r + Rc 1", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Consistent Phrase Error Rate", "sec_num": "3.2" }, { "text": "Note that CPER is equal to 1 -F-score. CPER penalizes incorrect or missing alignment links more severely than AER. While computing AER, an incorrect alignment link reduces the number of correct alignment links by 1, affecting precision and recall slightly. Similarly, if there is a missing link, only the recall is reduced slightly. However, when computing CPER, an incorrect or missing alignment link might result in more than one phrase pair being eliminated from or added to the set of phrases. Thus, the impact is more severe on both precision and recall. In Figure 1 , the first box represents a manual alignment and the other two represent automated alignments A. In the case of a missing alignment link ( Figure 1b ), P A includes 9 valid phrases. For this alignment, AER = 1 \u2212 (2 \u00d7 2/2 \u00d7 2/3)/(2/2 + 2/3) = 0.2 and CP ER = 1 \u2212 (2 \u00d7 5/9 \u00d7 5/6)/(5/9 + 5/6) = 0.33. In the case of an incorrect alignment link ( Figure 1c ), P A includes only 2 valid phrases, which results in a higher CPER (1 \u2212 (2 \u00d7 2/2 \u00d7 2/6)/(2/2 + 2/6) = 0.49) but a lower AER (1 \u2212 (2 \u00d7 3/4 \u00d7 3/3)/(3/4 + 3/3) = 0.14). Table 3 presents the CPER values on two different language pairs, using 2 different maximum phrase lengths. For both maximum phrase lengths, S A and S B yield the lowest CPER. For all 5 alignments-in both languages-CPER increases as the length of the phrase increases. For all alignments except S I , this amount of increase is nearly the same on both languages. Since S I contains very few alignment points, the number of generated phrases dramatically increases, yielding poor precision and CPER as the maximum phrase length increases.", "cite_spans": [], "ref_spans": [ { "start": 563, "end": 571, "text": "Figure 1", "ref_id": "FIGREF0" }, { "start": 712, "end": 721, "text": "Figure 1b", "ref_id": "FIGREF0" }, { "start": 916, "end": 925, "text": "Figure 1c", "ref_id": "FIGREF0" }, { "start": 1094, "end": 1101, "text": "Table 3", "ref_id": "TABREF5" } ], "eq_spans": [], "section": "Consistent Phrase Error Rate", "sec_num": "3.2" }, { "text": "We now move from intrinsic measurement to extrinsic measurement using an off-the-shelf phrasebased MT system Pharaoh (Koehn, 2004) . Our goal is to identify the characteristics of alignments that change MT behavior and the types of changes induced by these characteristics.", "cite_spans": [ { "start": 117, "end": 130, "text": "(Koehn, 2004)", "ref_id": "BIBREF8" } ], "ref_spans": [], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "All MT system components were kept the same in our experiments except for the component that generates a phrase table from a given alignment. We used the corpora presented in Table 1 to train the MT system. The phrases were scored using translation probabilities and lexical weights in two directions and a phrase penalty score. We also use a language model, a distortion model and a word penalty feature for MT.", "cite_spans": [], "ref_spans": [ { "start": 175, "end": 182, "text": "Table 1", "ref_id": "TABREF1" } ], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "We measure the impact of different alignments on Pharaoh using three different settings:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "1. Different maximum phrase length, 2. Different sizes of training data, and 3. Different lexical weighting.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "For maximum phrase length, we used 3 (based on what was suggested by (Koehn et al., 2003) and 7 (the default maximum phrase length in Pharaoh).", "cite_spans": [ { "start": 69, "end": 89, "text": "(Koehn et al., 2003)", "ref_id": "BIBREF7" } ], "ref_spans": [], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "For lexical weighting, we used the original weighting scheme employed in Pharaoh and a modified version. We realized that the publiclyavailable implementation of Pharaoh computes the lexical weights only for non-NULL alignment links. As a consequence, loose phrases containing NULL-aligned words along their edges receive the same lexical weighting as tight phrases without NULL-aligned words along the edges. We therefore adopted a modified weighting scheme following (Koehn et al., 2003) , which incorporates NULL alignments.", "cite_spans": [ { "start": 469, "end": 489, "text": "(Koehn et al., 2003)", "ref_id": "BIBREF7" } ], "ref_spans": [], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "MT output was evaluated using the standard evaluation metric BLEU (Papineni et al., 2002) . 2 The parameters of the MT System were optimized for BLEU metric on NIST MTEval'2002 test sets using minimum error rate training (Och, 2003) , and the systems were tested on NIST MTEval'2003 test sets for both languages. 2 We used the NIST script (version 11a) for BLEU with its default settings: case-insensitive matching of n-grams up to n = 4, and the shortest reference sentence for the brevity penalty. The words that were not translated during decoding were deleted from the MT output before running the BLEU script.", "cite_spans": [ { "start": 66, "end": 89, "text": "(Papineni et al., 2002)", "ref_id": "BIBREF16" }, { "start": 92, "end": 93, "text": "2", "ref_id": null }, { "start": 221, "end": 232, "text": "(Och, 2003)", "ref_id": "BIBREF15" }, { "start": 313, "end": 314, "text": "2", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "The SRI Language Modeling Toolkit was used to train a trigram model with modified Kneser-Ney smoothing on 155M words of English newswire text, mostly from the Xinhua portion of the Gigaword corpus. During decoding, the number of English phrases per FL phrase was limited to 100 and phrase distortion was limited to 4. Table 4 presents the BLEU scores for Pharaoh runs on Chinese with five different alignments using different settings for maximum phrase length (3 vs. 7), size of training data (107K vs. 241K), and lexical weighting (original vs. modified). 3 The modified lexical weighting yields huge improvements when the alignment leaves several words unaligned: the BLEU score for S A goes from 24.26 to 25.31 and the BLEU score for S B goes from 23.91 to 25.38. In contrast, when the alignments contain a high number of alignment links (e.g., S U and S G ), modifying lexical weighting does not bring significant improvements because the number of phrases containing unaligned words is relatively low. Increasing the phrase length increases the BLEU scores for all systems by nearly 0.7 points and increasing the size of the training data increases the BLEU scores by 1.5-2 points for all systems. For all settings, S U yields the lowest BLEU scores while S B clearly outperforms the others. Table 5 presents BLEU scores for Pharaoh runs on 5 different alignments on English-Arabic, using different settings for lexical weighting and maximum phrase lengths. 4 Using the original lexical weighting, S A and S B perform better than the others while S U and S I yield the worst results. Modifying the lexical weighting leads to slight reductions in BLEU scores for S U and S G , but improves the scores for the other 3 alignments significantly. Finally, increasing the maximum phrase length to 7 leads to additional improvements in BLEU scores, where S G and S U benefit nearly 2 BLEU points. As in English-Chinese, the worst BLEU scores are obtained by S U while the best scores are produced by S B .", "cite_spans": [ { "start": 558, "end": 559, "text": "3", "ref_id": null } ], "ref_spans": [ { "start": 318, "end": 325, "text": "Table 4", "ref_id": "TABREF7" }, { "start": 1298, "end": 1305, "text": "Table 5", "ref_id": "TABREF8" } ], "eq_spans": [], "section": "Evaluating Alignments within MT", "sec_num": "4" }, { "text": "As we see from the tables, the relation between intrinsic alignment measures (AER and CPER) and the corresponding BLEU scores varies, depending on the language, lexical weighting, maximum phrase length, and training data size. For example, using a modified lexical weighting, the systems are ranked according to their BLEU scores as follows: S B , S A , S G , S I , S U -an ordering that differs from that of AER but is identical to that of CPER (with a phrase length of 3) for Chinese. On the other hand, in Arabic, both AER and CPER provide a slightly different ranking from that of BLEU, with S G and S I swapping places.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "BLEU Score Comparison", "sec_num": "4.1" }, { "text": "To demonstrate how alignment-related components of the MT system might change the translation quality significantly, we did an additional experiment to compare different techniques for extracting phrases from a given alignment. Specifically, we are comparing two techniques for phrase extraction: 1. Loose phrases (the original 'consistent phrase extraction' method) 2. Tight phrases (the set of phrases where the first/last words on each side are forced to align to some word in the phrase pair) Using tight phrases penalizes alignments with many unaligned words, whereas using loose phrases rewards them. Our goal is to compare the performance of precision-oriented vs. recalloriented alignments when we allow only tight phrases in the phrase extraction step. To simplify things, we used only 2 alignments: S G , the best recall-oriented alignment, and S B , the best precision-oriented alignment. For this experiment, we used modified lexical weighting and a maximum phrase length of 7. Table 6 presents the BLEU scores for S G and S B using two different phrase extraction techniques on English-Chinese and English-Arabic. In both languages, S B outperforms S G significantly when loose phrases are used. However, when we use only tight phrases, the performance of S B gets significantly worse (3.5 to 6.5 BLEU-score reduction in comparison to loose phrases). The performance of S G also gets worse but the degree of BLEUscore reduction is less than that of S B . Overall S G performs better than S B with tight phrases; for English-Arabic, the difference between the two systems is more than 3 BLEU points. Note that, as before, the relation between the alignment measures and the BLEU scores varies, this time depending on whether loose phrases or tight phrases are used: both CPER and AER track the BLEU rankings for loose (but not for tight) phrases.", "cite_spans": [], "ref_spans": [ { "start": 990, "end": 997, "text": "Table 6", "ref_id": "TABREF10" } ], "eq_spans": [], "section": "Tight vs. Loose Phrases", "sec_num": "4.2" }, { "text": "This suggests that changing alignment-related components of the system (i.e., phrase extraction and phrase scoring) influences the overall translation quality significantly for a particular alignment. Therefore, when comparing two alignments in the context of a MT system, it is important to take the alignment characteristics into account. For instance, alignments with many unaligned words are severely penalized when using tight phrases.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Tight vs. Loose Phrases", "sec_num": "4.2" }, { "text": "We analyzed the percentage of words left untranslated during decoding. Figure 2 shows the percentage of untranslated words in the FL using the Chinese and Arabic NIST MTEval'2003 test sets.", "cite_spans": [], "ref_spans": [ { "start": 71, "end": 79, "text": "Figure 2", "ref_id": "FIGREF1" } ], "eq_spans": [], "section": "Untranslated Words", "sec_num": "4.3" }, { "text": "On English-Chinese data (using all four settings given in Table 4 ) S U and S G yield the highest percentage of untranslated words while S I produces the lowest percentage of untranslated words. S A and S B leave about 2% of the FL words phrases without translating them. Increasing the training data size reduces the percentage of untranslated words by nearly half with all five alignments. No significant impact on untranslated words is observed from modifying the lexical weights and changing the phrase length.", "cite_spans": [], "ref_spans": [ { "start": 58, "end": 65, "text": "Table 4", "ref_id": "TABREF7" } ], "eq_spans": [], "section": "Untranslated Words", "sec_num": "4.3" }, { "text": "On English-Arabic data, all alignments result in higher percentages of untranslated words than English-Chinese, most likely due to data sparsity. As in Chinese-to-English translation, S U is the worst and S B is the best. S I behaves quite differently, leaving nearly 7% of the words untranslated-an indicator of why it produces a higher BLEU score on Chinese but a lower score on Arabic compared to other alignments.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Untranslated Words", "sec_num": "4.3" }, { "text": "This section presents several experiments to analyze how different alignments affect the size of the generated phrase tables, the distribution of the phrases that are used in decoding, and the coverage of the test set with the generated phrase tables. Tables The major impact of using different alignments in a phrase-based MT system is that each one results in a different phrase table. Table 7 presents the number of phrases that are extracted from five alignments using two different maximum phrase lengths (3 vs. 7) in two languages, after filtering the phrase table for MTEval'2003 test set. The size of the phrase table increases dramatically as the number of links in the initial alignment gets smaller. As a result, for both languages, S U and S G yield a much smaller Chinese Arabic Alignment MPL=3 MPL=7 MPL=3 MPL=7 SU 106 122 32 38 SG 161 181 48 55 SI 1331 3498 377 984 SA 954 1856 297 594 SB 876 1624 262 486 Table 7 : Number of Phrases in the Phrase Table Filtered for MTEval '2003 Test Sets (in thousands) phrase table than the other three alignments. As the maximum phrase length increases, the size of the phrase table gets bigger for all alignments; however, the growth of the table is more significant for precision-oriented alignments due to the high number of unaligned words.", "cite_spans": [ { "start": 1018, "end": 1048, "text": "'2003 Test Sets (in thousands)", "ref_id": null } ], "ref_spans": [ { "start": 252, "end": 275, "text": "Tables The major impact", "ref_id": null }, { "start": 388, "end": 395, "text": "Table 7", "ref_id": null }, { "start": 777, "end": 957, "text": "Chinese Arabic Alignment MPL=3 MPL=7 MPL=3 MPL=7 SU 106 122 32 38 SG 161 181 48 55 SI 1331 3498 377 984 SA 954 1856 297 594 SB 876 1624 262 486 Table 7", "ref_id": "TABREF1" }, { "start": 992, "end": 1010, "text": "Table Filtered for", "ref_id": null } ], "eq_spans": [], "section": "Analysis of Phrase Tables", "sec_num": "4.4" }, { "text": "Distribution of Phrases To investigate how the decoder chooses phrases of different lengths, we analyzed the distribution of the phrases in the filtered phrase table and the phrases that were used to decode Chinese MTEval'2003 test set. 5 For the remaining experiments in the paper, we use modified lexical weighting, a maximum phrase length of 7, and 107K sentence pairs for training. The top row in Figure 3 shows the distribution of the phrases generated by the five alignments (using a maximum phrase length of 7) according to their length. The \"j-i\" designators correspond to the phrase pairs with j FL words and i English words. For S U and S G , the majority of the phrases contain only one FL word, and the percentage of the phrases with more than 2 FL words is less than 18%. For the other three alignments, however, the distribution of the phrases is almost inverted. For S I , nearly 62% of the phrases contain more than 3 words on either FL or English side; for S A and S B , this percentage is around 45-50%.", "cite_spans": [], "ref_spans": [ { "start": 401, "end": 409, "text": "Figure 3", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "Size of Phrase", "sec_num": null }, { "text": "Given the completely different phrase distribution, the most obvious question is whether the longer phrases generated by S I , S A and S B are actually used in decoding. In order to investigate this, we did an analysis of the phrases used to decode the same test set.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Size of Phrase", "sec_num": null }, { "text": "The bottom row of Figure 3 shows the percentage of phrases used to decode the Chinese MTEval'2003 test set. The distribution of the actual phrases used in decoding is completely the reverse of the distribution of the phrases in the entire filtered table. For all five alignments, the majority of the used phrases is one-to-one (between 50-65% of the total number of phrases used in decoding). S I , S A and S B use the other phrase pairs (particularly 1-to-2 phrases) more than S U and S G .", "cite_spans": [], "ref_spans": [ { "start": 18, "end": 26, "text": "Figure 3", "ref_id": "FIGREF2" } ], "eq_spans": [], "section": "Size of Phrase", "sec_num": null }, { "text": "Note that S I , S A and S B use only a small portion of the phrases with more than 3 words although the majority of the phrase table contains phrases with more than 3 words on one side. It is surprising that the inclusion of phrase pairs with more than 3 words in the search space increases the BLEU score although the majority of the phrases used in decoding is mostly one-to-one.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Size of Phrase", "sec_num": null }, { "text": "We also investigated the number and length of phrases that are used to decode the given test set for different alignments. Table 8 presents the average number of English and FL words in the phrases used in decoding Chinese MTEval'2003 test set. The decoder uses fewer phrases with S I , S A and S B than for the other two, thus yielding a higher number of FL words per phrase. The number of English words per phrase is also higher for these three systems than the other two.", "cite_spans": [], "ref_spans": [ { "start": 123, "end": 130, "text": "Table 8", "ref_id": "TABREF12" } ], "eq_spans": [], "section": "Length of the Phrases used in Decoding", "sec_num": null }, { "text": "Coverage of the Test Set Finally, we examine the coverage of a test set using phrases of a specific length in the phrase table. Table 9 reveals that the coverage of the test set is higher for precision-oriented alignments than recall-oriented alignments for all different lengths of the phrases. For instance, S I , S A , and S B cover nearly 75% of the corpus using only phrases with 2 FL words, and nearly 36% of the corpus using phrases with 3 FL words. This suggests that recalloriented alignments fail to catch a significant number of phrases that would be useful to decode this test set, and precision-oriented alignments would yield potentially more useful phrases.", "cite_spans": [], "ref_spans": [ { "start": 128, "end": 135, "text": "Table 9", "ref_id": "TABREF11" } ], "eq_spans": [], "section": "Length of the Phrases used in Decoding", "sec_num": null }, { "text": "Since precision-oriented alignments make a higher number of longer phrases available to the decoder (based on the coverage of phrases presented in Table 9 ), they are used more during decoding. Consequently, the major difference between the alignments is the coverage of the phrases extracted from different alignments. The more the phrase table covers the test set, the more the longer phrases are used during decoding, and precision-oriented alignments are better at generating high-coverage phrases than recall-oriented alignments.", "cite_spans": [], "ref_spans": [ { "start": 147, "end": 154, "text": "Table 9", "ref_id": "TABREF11" } ], "eq_spans": [], "section": "Length of the Phrases used in Decoding", "sec_num": null }, { "text": "This paper investigated how different alignments change the behavior of phrase-based MT. We showed that AER is a poor indicator of MT performance because it penalizes incorrect links less than is reflected in the corresponding phrasebased MT. During phrase-based MT, an incorrect alignment link might prevent extraction of several phrases, but the number of phrases affected by that link depends on the context.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusions and Future Work", "sec_num": "5" }, { "text": "We designed CPER, a new phrase-oriented metric that is more informative than AER when the alignments are used in a phrase-based MT system because it is an indicator of how the set of phrases differ from one alignment to the next according to a pre-specified maximum phrase length.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusions and Future Work", "sec_num": "5" }, { "text": "Even with refined evaluation metrics (including CPER), we found it difficult to assess the impact of alignment on MT performance because word alignment is not the only factor that affects the choice of the correct words (or phrases) during decoding. We empirically showed that different phrase extraction techniques result in better MT output for certain alignments but the MT performance gets worse for other alignments. Similarly, adjusting the scores assigned to the phrases makes a significant difference for certain alignments while it has no impact on some others. Consequently, when comparing two BLEU scores, it is difficult to determine whether the alignments are bad to start with or the set of extracted phrases is bad or the phrases extracted from the alignments are assigned bad scores. This suggests that finding a direct correlation between AER (or even CPER) and the automated MT metrics is infeasible.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusions and Future Work", "sec_num": "5" }, { "text": "We demonstrated that recall-oriented alignment methods yield smaller phrase tables and a higher number of untranslated words when compared to precision-oriented methods. We also showed that the phrases extracted from recall-oriented alignments cover a smaller portion of a given test set when compared to precision-oriented alignments. Finally, we showed that the decoder with recalloriented alignments uses shorter phrases more frequently as a result of unavailability of longer phrases that are extracted. Future work will involve an investigation into how the phrase extraction and scoring should be adjusted to take the nature of the alignment into account and how the phrase-table size might be reduced without sacrificing the MT output quality.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Conclusions and Future Work", "sec_num": "5" }, { "text": "We could not run SB on the larger corpus because of the lack of required inputs.4 Due to lack of additional training data, we could not do experiments using different sizes of training data on English-Arabic.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null }, { "text": "Due to lack of space, we will present results on Chinese-English only in the rest of this paper but the Arabic-English results show the same trends.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "Neuralign: Combining word alignments using neural networks", "authors": [ { "first": "F", "middle": [], "last": "Necip", "suffix": "" }, { "first": "Bonnie", "middle": [ "J" ], "last": "Ayan", "suffix": "" }, { "first": "Christof", "middle": [], "last": "Dorr", "suffix": "" }, { "first": "", "middle": [], "last": "Monz", "suffix": "" } ], "year": 2005, "venue": "Proceedings of EMNLP'2005", "volume": "", "issue": "", "pages": "65--72", "other_ids": {}, "num": null, "urls": [], "raw_text": "Necip F. Ayan, Bonnie J. Dorr, and Christof Monz. 2005. Neuralign: Combining word alignments using neural net- works. In Proceedings of EMNLP'2005, pages 65-72.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "Meteor: An automatic metric for MT evaluation with improved correlation with human judgments", "authors": [ { "first": "Stanjeev", "middle": [], "last": "Banerjee", "suffix": "" }, { "first": "Alon", "middle": [], "last": "Lavie", "suffix": "" } ], "year": 2005, "venue": "Proceedings of Workshop on Intrinsic and Extrinsic Evaluation Measures for MT and/or Summarization at ACL-2005", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Stanjeev Banerjee and Alon Lavie. 2005. Meteor: An au- tomatic metric for MT evaluation with improved corre- lation with human judgments. In Proceedings of Work- shop on Intrinsic and Extrinsic Evaluation Measures for MT and/or Summarization at ACL-2005.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "The mathematics of statistical machine translation: Parameter estimation", "authors": [ { "first": "F", "middle": [], "last": "Peter", "suffix": "" }, { "first": "Stephan", "middle": [ "A" ], "last": "Brown", "suffix": "" }, { "first": "Robert", "middle": [ "L" ], "last": "Della Pietra", "suffix": "" }, { "first": "", "middle": [], "last": "Mercer", "suffix": "" } ], "year": 1993, "venue": "Computational Linguistics", "volume": "19", "issue": "2", "pages": "263--311", "other_ids": {}, "num": null, "urls": [], "raw_text": "Peter F. Brown, Stephan A. Della Pietra, and Robert L. Mer- cer. 1993. The mathematics of statistical machine trans- lation: Parameter estimation. Computational Linguistics, 19(2):263-311.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Statistical machine translation with word-and sentence-aligned parallel corpora", "authors": [ { "first": "Chris", "middle": [], "last": "Callison", "suffix": "" }, { "first": "-", "middle": [], "last": "Burch", "suffix": "" }, { "first": "David", "middle": [], "last": "Talbot", "suffix": "" }, { "first": "Miles", "middle": [], "last": "Osborne", "suffix": "" } ], "year": 2004, "venue": "Proceedings of ACL", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Chris Callison-Burch, David Talbot, and Miles Osborne. 2004. Statistical machine translation with word-and sentence-aligned parallel corpora. In Proceedings of ACL'2004.", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "A hierarchical phrase-based model for statistical machine translation", "authors": [ { "first": "David", "middle": [], "last": "Chiang", "suffix": "" } ], "year": 2005, "venue": "Proceedings of ACL'2005", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "David Chiang. 2005. A hierarchical phrase-based model for statistical machine translation. In Proceedings of ACL'2005.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Aligning words using matrix factorisation", "authors": [ { "first": "Cyril", "middle": [], "last": "Goutte", "suffix": "" }, { "first": "Kenji", "middle": [], "last": "Yamada", "suffix": "" }, { "first": "Eric", "middle": [], "last": "Gaussier", "suffix": "" } ], "year": 2004, "venue": "Proceedings of ACL'2004", "volume": "", "issue": "", "pages": "502--509", "other_ids": {}, "num": null, "urls": [], "raw_text": "Cyril Goutte, Kenji Yamada, and Eric Gaussier. 2004. Align- ing words using matrix factorisation. In Proceedings of ACL'2004, pages 502-509.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "A maximum entropy word aligner for arabic-english machine translation", "authors": [ { "first": "Abraham", "middle": [], "last": "Ittycheriah", "suffix": "" }, { "first": "Salim", "middle": [], "last": "Roukos", "suffix": "" } ], "year": 2005, "venue": "Proceedings of EMNLP", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Abraham Ittycheriah and Salim Roukos. 2005. A maximum entropy word aligner for arabic-english machine transla- tion. In Proceedings of EMNLP'2005.", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "Statistical phrase-based translation", "authors": [ { "first": "Philipp", "middle": [], "last": "Koehn", "suffix": "" }, { "first": "Franz", "middle": [ "J" ], "last": "Och", "suffix": "" }, { "first": "Daniel", "middle": [], "last": "Marcu", "suffix": "" } ], "year": 2003, "venue": "Proceedings of HLT-NAACL", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Philipp Koehn, Franz J. Och, and Daniel Marcu. 2003. Sta- tistical phrase-based translation. In Proceedings of HLT- NAACL'2003.", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "Pharaoh: A beam search decoder for phrase-based statistical machine translation", "authors": [ { "first": "Philipp", "middle": [], "last": "Koehn", "suffix": "" } ], "year": 2004, "venue": "Proceedings of AMTA", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Philipp Koehn. 2004. Pharaoh: A beam search decoder for phrase-based statistical machine translation. In Proceed- ings of AMTA'2004.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "A phrase-based, joint probability model for statistical machine translation", "authors": [ { "first": "Daniel", "middle": [], "last": "Marcu", "suffix": "" }, { "first": "William", "middle": [], "last": "Wong", "suffix": "" } ], "year": 2002, "venue": "Proceedings of EMNLP", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Daniel Marcu and William Wong. 2002. A phrase-based, joint probability model for statistical machine translation. In Proceedings of EMNLP'2002.", "links": null }, "BIBREF10": { "ref_id": "b10", "title": "Models of translational equivalence among words", "authors": [ { "first": "I", "middle": [], "last": "", "suffix": "" }, { "first": "Dan", "middle": [], "last": "Melamed", "suffix": "" } ], "year": 2000, "venue": "Computational Linguistics", "volume": "26", "issue": "2", "pages": "221--249", "other_ids": {}, "num": null, "urls": [], "raw_text": "I. Dan Melamed. 2000. Models of translational equivalence among words. Computational Linguistics, 26(2):221- 249.", "links": null }, "BIBREF11": { "ref_id": "b11", "title": "A discriminative framework for bilingual word alignment", "authors": [ { "first": "Robert", "middle": [ "C" ], "last": "Moore", "suffix": "" } ], "year": 2005, "venue": "Proceedings of EMNLP'2005", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Robert C. Moore. 2005. A discriminative frame- work for bilingual word alignment. In Proceedings of EMNLP'2005.", "links": null }, "BIBREF12": { "ref_id": "b12", "title": "A comparison of alignment models for statistical machine translation", "authors": [ { "first": "J", "middle": [], "last": "Franz", "suffix": "" }, { "first": "Hermann", "middle": [], "last": "Och", "suffix": "" }, { "first": "", "middle": [], "last": "Ney", "suffix": "" } ], "year": 2000, "venue": "Proceedings of COLING'", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Franz J. Och and Hermann Ney. 2000. A comparison of alignment models for statistical machine translation. In Proceedings of COLING'2000.", "links": null }, "BIBREF13": { "ref_id": "b13", "title": "GIZA++: Training of statistical translation models", "authors": [ { "first": "Franz", "middle": [ "J" ], "last": "Och", "suffix": "" } ], "year": 2000, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Franz J. Och. 2000b. GIZA++: Training of statistical transla- tion models. Technical report, RWTH Aachen, University of Technology.", "links": null }, "BIBREF14": { "ref_id": "b14", "title": "A systematic comparison of various statistical alignment models", "authors": [ { "first": "J", "middle": [], "last": "Franz", "suffix": "" }, { "first": "Hermann", "middle": [], "last": "Och", "suffix": "" }, { "first": "", "middle": [], "last": "Ney", "suffix": "" } ], "year": 2003, "venue": "Computational Linguistics", "volume": "29", "issue": "1", "pages": "9--51", "other_ids": {}, "num": null, "urls": [], "raw_text": "Franz J. Och and Hermann Ney. 2003. A systematic com- parison of various statistical alignment models. Computa- tional Linguistics, 29(1):9-51, March.", "links": null }, "BIBREF15": { "ref_id": "b15", "title": "Minimum error rate training in statistical machine translation", "authors": [ { "first": "Franz", "middle": [ "J" ], "last": "Och", "suffix": "" } ], "year": 2003, "venue": "Proceedings of ACL", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Franz J. Och. 2003. Minimum error rate training in statistical machine translation. In Proceedings of ACL'2003.", "links": null }, "BIBREF16": { "ref_id": "b16", "title": "BLEU: a method for automatic evaluation of machine translation", "authors": [ { "first": "Kishore", "middle": [], "last": "Papineni", "suffix": "" }, { "first": "Salim", "middle": [], "last": "Roukos", "suffix": "" }, { "first": "Todd", "middle": [], "last": "Ward", "suffix": "" }, { "first": "Wei-Jing", "middle": [], "last": "Zhu", "suffix": "" } ], "year": 2002, "venue": "Proceedings of ACL", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. BLEU: a method for automatic evaluation of machine translation. In Proceedings of ACL'2002.", "links": null }, "BIBREF17": { "ref_id": "b17", "title": "A discriminative matching approach to word alignment", "authors": [ { "first": "Ben", "middle": [], "last": "Taskar", "suffix": "" }, { "first": "Simon", "middle": [], "last": "Lacoste-Julien", "suffix": "" }, { "first": "Dan", "middle": [], "last": "Klein", "suffix": "" } ], "year": 2005, "venue": "Proceedings of EMNLP", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Ben Taskar, Simon Lacoste-Julien, and Dan Klein. 2005. A discriminative matching approach to word alignment. In Proceedings of EMNLP'2005.", "links": null }, "BIBREF18": { "ref_id": "b18", "title": "HMM-based word alignment in statistical translation", "authors": [ { "first": "Stefan", "middle": [], "last": "Vogel", "suffix": "" }, { "first": "Hermann", "middle": [], "last": "Ney", "suffix": "" }, { "first": "Christoph", "middle": [], "last": "Tillmann", "suffix": "" } ], "year": 1996, "venue": "Proceedings of COLING'1996", "volume": "", "issue": "", "pages": "836--841", "other_ids": {}, "num": null, "urls": [], "raw_text": "Stefan Vogel, Hermann Ney, and Christoph Tillmann. 1996. HMM-based word alignment in statistical translation. In Proceedings of COLING'1996, pages 836-841.", "links": null } }, "ref_entries": { "FIGREF0": { "type_str": "figure", "text": "Sample phrases that are generated from a human alignment and an automated alignment: Gray cells show the alignment links, and rectangles show the possible phrases.", "num": null, "uris": null }, "FIGREF1": { "type_str": "figure", "text": "Percentage of untranslated words out of the total number of FL words", "num": null, "uris": null }, "FIGREF2": { "type_str": "figure", "text": "Distribution of the phrases in the phrase table filtered for Chinese MTEval'2003 test set (top row) and the phrases used in decoding the same test set (bottom row) according to their lengths", "num": null, "uris": null }, "TABREF1": { "text": "Test and Training Data Used for Experiments", "content": "", "num": null, "type_str": "table", "html": null }, "TABREF3": { "text": "", "content": "
", "num": null, "type_str": "table", "html": null }, "TABREF5": { "text": "Consistent Phrase Error Rates with Maximum", "content": "
Phrase Lengths of 3 and 7
", "num": null, "type_str": "table", "html": null }, "TABREF7": { "text": "BLEU Scores on English-Chinese with Different Lexical Weightings, Maximum Phrase Lengths and Training Data", "content": "
LW=Org LW=Mod LW=Mod
AlignmentMPL=3MPL=3MPL=7
SU41.9741.7243.50
SG44.0643.8245.78
SI42.2942.7643.88
SA44.4945.2346.06
SB44.9245.3946.66
", "num": null, "type_str": "table", "html": null }, "TABREF8": { "text": "", "content": "
: BLEU Scores on English-Arabic with Different
Lexical Weightings and Maximum Phrase Lengths
", "num": null, "type_str": "table", "html": null }, "TABREF10": { "text": "BLEU Scores with Loose vs. Tight Phrases", "content": "", "num": null, "type_str": "table", "html": null }, "TABREF11": { "text": "", "content": "
Alignment |Eng| |FL|
SU1.39 1.28
SG1.45 1.33
SI1.51 1.55
SA1.54 1.55
SB1.56 1.52
", "num": null, "type_str": "table", "html": null }, "TABREF12": { "text": "The average length of the phrases that are used in", "content": "
decoding Chinese MTEval'2003 test set
the coverage of the Chinese MTEval'2003 test set
(source side) using only phrases of a particular
length (from 1 to 7). For this experiment, we as-
sume that a word in the test set is covered if it is
part of a phrase pair that exists in the phrase table
(if a word is part of multiple phrases, it is counted
only once). Not surprisingly, using only phrases
with one FL word, more than 90% of the test set
can be covered for all 5 alignments. As the length
of the phrases increases, the coverage of the test
set decreases. For instance, using phrases with 5
FL words results in less than 5% coverage of the
test set.
Phrase Length (FL)
A1234567
SU 92.2 59.5 21.46.71.3 0.4 0.1
SG 95.5 64.4 24.97.41.6 0.5 0.3
SI97.8 75.8 38.0 13.8 4.6 1.9 1.2
SA 97.3 75.3 36.1 12.5 3.8 1.5 0.8
SB 97.5 74.8 35.7 12.4 4.2 1.8 0.9
", "num": null, "type_str": "table", "html": null }, "TABREF13": { "text": "Coverage of Chinese MTEval'2003 Test Set UsingPhrases with a Specific Length on FL side (in percentages)", "content": "", "num": null, "type_str": "table", "html": null } } } }