| { |
| "paper_id": "2021", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T13:27:46.634401Z" |
| }, |
| "title": "Investigating non lexical markers of the language of schizophrenia in spontaneous conversations", |
| "authors": [ |
| { |
| "first": "Chuyuan", |
| "middle": [], |
| "last": "Li", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "CNRS", |
| "location": { |
| "postCode": "F-54000", |
| "settlement": "Nancy", |
| "region": "Inria, LORIA", |
| "country": "France" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Maxime", |
| "middle": [], |
| "last": "Amblard", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "CNRS", |
| "location": { |
| "postCode": "F-54000", |
| "settlement": "Nancy", |
| "region": "Inria, LORIA", |
| "country": "France" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Chlo\u00e9", |
| "middle": [], |
| "last": "Braud", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "IRIT", |
| "institution": "ANITI", |
| "location": { |
| "settlement": "Toulouse", |
| "country": "France" |
| } |
| }, |
| "email": "2chloe.braud@irit.fr" |
| }, |
| { |
| "first": "Caroline", |
| "middle": [], |
| "last": "Demily", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "UMR 5229", |
| "institution": "Univerist\u00e9", |
| "location": { |
| "addrLine": "Lyon 1", |
| "settlement": "Lyon", |
| "country": "France" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Nicolas", |
| "middle": [], |
| "last": "Franck", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "UMR 5229", |
| "institution": "Univerist\u00e9", |
| "location": { |
| "addrLine": "Lyon 1", |
| "settlement": "Lyon", |
| "country": "France" |
| } |
| }, |
| "email": "" |
| }, |
| { |
| "first": "Michel", |
| "middle": [], |
| "last": "Musiol", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "CNRS", |
| "location": { |
| "postCode": "F-54000", |
| "settlement": "Nancy", |
| "region": "Inria, LORIA", |
| "country": "France" |
| } |
| }, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "We investigate linguistic markers associated with schizophrenia in clinical conversations by detecting predictive features among Frenchspeaking patients. Dealing with humanhuman dialogues makes for a realistic situation, but it calls for strategies to represent the context and face data sparsity. We compare different approaches for data representation-from individual speech turns to entire conversations-, and data modeling, using lexical, morphological, syntactic, and discourse features, dimensions presumed to be tightly connected to the language of schizophrenia. Previous English models were mostly lexical and reached high performance, here replicated (93.7% acc.). However, our analysis reveals that these models are heavily biased, which probably concerns most datasets on this task. Our new delexicalized models are more general and robust, with the best accuracy score at 77.9%.", |
| "pdf_parse": { |
| "paper_id": "2021", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "We investigate linguistic markers associated with schizophrenia in clinical conversations by detecting predictive features among Frenchspeaking patients. Dealing with humanhuman dialogues makes for a realistic situation, but it calls for strategies to represent the context and face data sparsity. We compare different approaches for data representation-from individual speech turns to entire conversations-, and data modeling, using lexical, morphological, syntactic, and discourse features, dimensions presumed to be tightly connected to the language of schizophrenia. Previous English models were mostly lexical and reached high performance, here replicated (93.7% acc.). However, our analysis reveals that these models are heavily biased, which probably concerns most datasets on this task. Our new delexicalized models are more general and robust, with the best accuracy score at 77.9%.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "Schizophrenia is defined as a severe mental illness (APA, 2015) that comes with varied symptoms, ranging from delirium to hallucinations. Among these symptoms, there are language disorders, especially the so-called positive thought disorder (i.e., disorganized language output such as derailment and tangentiality) 1 and negative thought disorder 2 (Kuperberg, 2010). Schizophrenia affects about 1% of the world's adult population, with cognitive troubles for 70-80% of the patients (Potvin et al., 2017) . Since the symptoms often affect language skills, several studies proposed using NLP techniques on patients' productions (Section 2) to identify what is affected in language, thus understand better the 1 Derailment: spontaneous speech that tends to slip off track. Tangentiality: reply to a question in an oblique or irrelevant manner.", |
| "cite_spans": [ |
| { |
| "start": 483, |
| "end": 504, |
| "text": "(Potvin et al., 2017)", |
| "ref_id": "BIBREF35" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "2 Negative thought disorder are those of poverty of speech and language (known as alogia) and poverty of content. disease and its symptoms and how language works in general.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "In this paper, we explore linguistic markers of schizophrenia through feature exploration within a classification system. We do so on spontaneous dialogues in French where all the previous work was in English and most used social media data or monologues. Replicating state-of-the-art results allows us to confirm some previous findings of specific features of the language of schizophrenia.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Our study focuses on two aspects: carefully exploring data representations and investigating preliminary modeling of dialogues, both with scarce data. Using spontaneous conversations makes for a realistic scenario -the patient is merely talking with her clinician. However, representing dialogues is not easy: we restrict ourselves to patients' speech turns, and test varied context windows to tackle data sparsity. Additionally, we compare several representations and confirm that lexicon is a good indicator, making for high-performing models with at best 93.7% (acc.). Nevertheless, our analysis demonstrates that it probably corresponds to a bias in our data caused by the constraints imposed during the collection process. Most of the datasets are likely biased the same way. This analysis led us to delexicalized models while focusing on dimensions presumed to be affected in schizophrenia: morphosyntactic, syntactic, dialogue, and discourse information are therefore considered. Our best delexicalized model gets 77.9% (acc.) and shows the importance of morpho-syntactic information and high-level features in dialogue.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "When dealing with medical data, ethical questions arise. The diagnosis of schizophrenia is complex and relies on many indices. Automatic systems could provide psychiatrists with further clues, possibly alleviating the need for the patients to go through several cognitive tests, but this is a longreach goal. It is clear that the systems developed can not substitute for a human expert, as a diagnosis is a medical act. Moreover, linguistic clues, while crucial, have to be interpreted within the patient's social environment.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Contributions This study: (i) is the first in French, replicating English studies with comparable results with less data and resources; (ii) continues seminal work on schizophrenia detection in dialogues but with a focus on modeling and bias -two crucial issues for a task inherently data-scarce; (iii) reveals language features of schizophrenia, confirming psychologists' descriptions on the use of complex structures or the capacity to maintain conversation. 3", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Introduction", |
| "sec_num": "1" |
| }, |
| { |
| "text": "Psychiatrists rely on language and speech behavior as one of the main clues in psychiatric diagnosis (Ratana et al., 2019) . They found that these patients' speech tends to be less predictable (Salzinger et al., 1964 (Salzinger et al., , 1970 Salzinger, 1979) , with a poorer vocabulary (Salzinger and Hammer, 1963; Manschreck et al., 1991) . It has also been found that their productions tend to be more grammatically deviant (Hoffman and Sledge, 1988) and less syntactically complex than that of controls (Fraser et al., 1986; Morice and Ingram, 1982) . At the discourse level, they associate words within a larger context than controls (Maher et al., 2005) with often more diffuse associations (Chaika, 1974; Elvev\u00e5g et al., 2007) . They also present referential impairments -categorized as vagueness, missing information, or confusing reference (Rochester, 2013; Docherty et al., 1996) -, and specific discontinuities at the discourse level (Musiol and Trognon, 2000; Rebuschi et al., 2014) .", |
| "cite_spans": [ |
| { |
| "start": 101, |
| "end": 122, |
| "text": "(Ratana et al., 2019)", |
| "ref_id": "BIBREF36" |
| }, |
| { |
| "start": 193, |
| "end": 216, |
| "text": "(Salzinger et al., 1964", |
| "ref_id": null |
| }, |
| { |
| "start": 217, |
| "end": 242, |
| "text": "(Salzinger et al., , 1970", |
| "ref_id": null |
| }, |
| { |
| "start": 243, |
| "end": 259, |
| "text": "Salzinger, 1979)", |
| "ref_id": null |
| }, |
| { |
| "start": 287, |
| "end": 315, |
| "text": "(Salzinger and Hammer, 1963;", |
| "ref_id": null |
| }, |
| { |
| "start": 316, |
| "end": 340, |
| "text": "Manschreck et al., 1991)", |
| "ref_id": "BIBREF27" |
| }, |
| { |
| "start": 427, |
| "end": 453, |
| "text": "(Hoffman and Sledge, 1988)", |
| "ref_id": "BIBREF16" |
| }, |
| { |
| "start": 507, |
| "end": 528, |
| "text": "(Fraser et al., 1986;", |
| "ref_id": "BIBREF13" |
| }, |
| { |
| "start": 529, |
| "end": 553, |
| "text": "Morice and Ingram, 1982)", |
| "ref_id": "BIBREF29" |
| }, |
| { |
| "start": 639, |
| "end": 659, |
| "text": "(Maher et al., 2005)", |
| "ref_id": "BIBREF26" |
| }, |
| { |
| "start": 697, |
| "end": 711, |
| "text": "(Chaika, 1974;", |
| "ref_id": "BIBREF8" |
| }, |
| { |
| "start": 712, |
| "end": 733, |
| "text": "Elvev\u00e5g et al., 2007)", |
| "ref_id": "BIBREF11" |
| }, |
| { |
| "start": 849, |
| "end": 866, |
| "text": "(Rochester, 2013;", |
| "ref_id": "BIBREF38" |
| }, |
| { |
| "start": 867, |
| "end": 889, |
| "text": "Docherty et al., 1996)", |
| "ref_id": "BIBREF10" |
| }, |
| { |
| "start": 945, |
| "end": 971, |
| "text": "(Musiol and Trognon, 2000;", |
| "ref_id": "BIBREF30" |
| }, |
| { |
| "start": 972, |
| "end": 994, |
| "text": "Rebuschi et al., 2014)", |
| "ref_id": "BIBREF37" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "On the other hand, many researchers have used NLP methods to help to identify mental disorders, such as depression (Howes et al., 2014; Guntuku et al., 2019; Sekuli\u0107 and Strube, 2019) , posttraumatic stress disorder (Pedersen, 2015; Kleim et al., 2018) , suicide risk (Benton et al., 2017 ), Alzheimer's disease (Orimaye et al., 2014; Fraser et al., 2016) , and autism (Goodkind et al., 2018; Sakishita et al., 2019) .", |
| "cite_spans": [ |
| { |
| "start": 115, |
| "end": 135, |
| "text": "(Howes et al., 2014;", |
| "ref_id": "BIBREF20" |
| }, |
| { |
| "start": 136, |
| "end": 157, |
| "text": "Guntuku et al., 2019;", |
| "ref_id": "BIBREF15" |
| }, |
| { |
| "start": 158, |
| "end": 183, |
| "text": "Sekuli\u0107 and Strube, 2019)", |
| "ref_id": null |
| }, |
| { |
| "start": 216, |
| "end": 232, |
| "text": "(Pedersen, 2015;", |
| "ref_id": "BIBREF32" |
| }, |
| { |
| "start": 233, |
| "end": 252, |
| "text": "Kleim et al., 2018)", |
| "ref_id": "BIBREF24" |
| }, |
| { |
| "start": 268, |
| "end": 288, |
| "text": "(Benton et al., 2017", |
| "ref_id": "BIBREF4" |
| }, |
| { |
| "start": 312, |
| "end": 334, |
| "text": "(Orimaye et al., 2014;", |
| "ref_id": "BIBREF31" |
| }, |
| { |
| "start": 335, |
| "end": 355, |
| "text": "Fraser et al., 2016)", |
| "ref_id": "BIBREF12" |
| }, |
| { |
| "start": 369, |
| "end": 392, |
| "text": "(Goodkind et al., 2018;", |
| "ref_id": "BIBREF14" |
| }, |
| { |
| "start": 393, |
| "end": 416, |
| "text": "Sakishita et al., 2019)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "For schizophrenia, previous work has mainly focused on lexical information (Mitchell et al., 2015; Hong et al., 2012; Birnbaum et al., 2017; Xu et al., 2019) . Unlike ours, these studies rely on Linguistic Inquiry Word Count (LIWC) cate-gories (Pennebaker et al., 2001 ) -psycho-metrically validated lexicon mapping words to psychological concepts), Latent Dirichlet Allocation (LDA) (Blei et al., 2003) -inferring topics in each document, and Brown clustering (Brown et al., 1992) -grouping contextually similar words into the same cluster. However, most of these resources are only available in English.", |
| "cite_spans": [ |
| { |
| "start": 75, |
| "end": 98, |
| "text": "(Mitchell et al., 2015;", |
| "ref_id": "BIBREF28" |
| }, |
| { |
| "start": 99, |
| "end": 117, |
| "text": "Hong et al., 2012;", |
| "ref_id": "BIBREF17" |
| }, |
| { |
| "start": 118, |
| "end": 140, |
| "text": "Birnbaum et al., 2017;", |
| "ref_id": "BIBREF5" |
| }, |
| { |
| "start": 141, |
| "end": 157, |
| "text": "Xu et al., 2019)", |
| "ref_id": null |
| }, |
| { |
| "start": 244, |
| "end": 268, |
| "text": "(Pennebaker et al., 2001", |
| "ref_id": "BIBREF34" |
| }, |
| { |
| "start": 384, |
| "end": 403, |
| "text": "(Blei et al., 2003)", |
| "ref_id": "BIBREF6" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "More recent approaches considered syntactic, semantic, and sentiment information (Kayi et al., 2017; Allende-Cid et al., 2019) . Both studies show good performance with morpho-syntactic features, especially with Part-Of-Speech (POS) tags. 4 They were based on narrative texts (essays and tweets). We here demonstrate that some findings can generalize to spontaneous conversations. Amblard et al. (2020) proposed the first study on detecting schizophrenia patients from conversations, mostly limited to lexical features. Also, close to our work, Howes et al. (2012a Howes et al. ( ,b, 2013 investigated linguistic features in transcripts of conversations between patients and clinicians. The authors tried to predict patient satisfaction and adherence to treatment on the concatenation of speech turns of the patient. Inspired by the work of Howes et al. 2012b, we also use higher-level features (see Section 3) on real conversations but directly investigating a model of detecting patients with schizophrenia symptoms. Furthermore, we extend previous work by varying the length of dialogues and testing more complex features, including sequences of POS tags, finer tree representations, and dialogical information.", |
| "cite_spans": [ |
| { |
| "start": 81, |
| "end": 100, |
| "text": "(Kayi et al., 2017;", |
| "ref_id": "BIBREF23" |
| }, |
| { |
| "start": 101, |
| "end": 126, |
| "text": "Allende-Cid et al., 2019)", |
| "ref_id": "BIBREF0" |
| }, |
| { |
| "start": 239, |
| "end": 240, |
| "text": "4", |
| "ref_id": null |
| }, |
| { |
| "start": 381, |
| "end": 402, |
| "text": "Amblard et al. (2020)", |
| "ref_id": "BIBREF1" |
| }, |
| { |
| "start": 545, |
| "end": 564, |
| "text": "Howes et al. (2012a", |
| "ref_id": "BIBREF18" |
| }, |
| { |
| "start": 565, |
| "end": 588, |
| "text": "Howes et al. ( ,b, 2013", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Related work", |
| "sec_num": "2" |
| }, |
| { |
| "text": "Varying dialogue size: Our data are composed of 41 dialogues with 2, 811 words, and 268 speech turns on average (when limited to patients/controls). The clinician's speech turns are ignored in all dialogues to reduce their impact on classification, but further studies should also include the interaction. First, we concatenate all the speech turns of a patient/control (Full setting), thus making for a large document that contains the whole context. Since the documents are long, it could be hard for the system to find regularities, especially with only a few classification instances (i.e., 41). The opposite #Doc. #Speech T./doc. #Word/doc. option is to classify each speech turn individually (Indiv.): this leads to more instances (10, 319), but we lose the context of the neighboring speech turns. Moreover, the speech turns are of varied length with an average of 11 words; some of them contain too few words to be informative. The last option is in between: we use a window of at least n words (W-n), always going until the end of the current speech turn, to assess the possibility of identifying distinctive features already in smaller blocks of conversation. We test n \u2208 {128, 256, 512}, providing with middle representations (see Table 1 ). The number of instances is (resp.) 893, 443, and 209, with an average number of speech turns 11, 20, and 42. This configuration allows keeping some context without overwhelming the model.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 1242, |
| "end": 1249, |
| "text": "Table 1", |
| "ref_id": "TABREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Approach", |
| "sec_num": "3" |
| }, |
| { |
| "text": "Comparing representations: Existing work on schizophrenia language demonstrated the importance of lexical features. For French, as for many languages, we do not have access to a resource such as LIWC. We thus propose to simply include bag-of-words (bow) and n-grams (n \u2208 {2, 3}) to our models as a proxy for topic identification. Howes et al. (2012b) showed the importance of features specific to spontaneous dialogues that do involve lexicon but in a more generic way: OCR corresponds to Open Class Repair initiators (pardon?, huh?); Backchannel (BC) responses are phatic expressions (yeah, hum mm). To reflect text organization, we also include discourse features by extracting the forms (without disambiguation) corresponding to connectives (but, because, since) as identified in LexConn (Roze et al., 2012) .", |
| "cite_spans": [ |
| { |
| "start": 330, |
| "end": 350, |
| "text": "Howes et al. (2012b)", |
| "ref_id": "BIBREF21" |
| }, |
| { |
| "start": 791, |
| "end": 810, |
| "text": "(Roze et al., 2012)", |
| "ref_id": "BIBREF39" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Approach", |
| "sec_num": "3" |
| }, |
| { |
| "text": "Finally, we test the two following non-lexical features: Part-Of-Speech n-grams and treelets. Allende-Cid et al. (2019) demonstrated that POS tags are effective features. We also test for larger patterns with sequences, POS n-gram with n \u2208 {1, 2, 3}. Kayi et al. (2017) only used the dependencies as syntactic features. We extend to treelet features (Johannsen et al., 2015) pendency parse trees: 2-treelet corresponds to 2 tokens with a syntactic relation between a head and a dependent, e.g., 'VERB\u2192nsubj\u2192NOUN', and 3-treelet corresponds to 3 tokens with 2 syntactic relations: could be 1 head dominates 2 dependents or a chain of dependencies, e.g., 'PRON\u2190poss\u2190NOUN\u2190nsubj\u2190VERB'. See Figure 1 for an illustration.", |
| "cite_spans": [ |
| { |
| "start": 350, |
| "end": 374, |
| "text": "(Johannsen et al., 2015)", |
| "ref_id": "BIBREF22" |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 686, |
| "end": 694, |
| "text": "Figure 1", |
| "ref_id": "FIGREF0" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Approach", |
| "sec_num": "3" |
| }, |
| { |
| "text": "Data: Forty-one conversations between patients (18) or controls (23) and a psychologist come from SLAM project (Rebuschi et al., 2014; Amblard et al., 2015) . The transcripts are standardized and follow a transcription guide. The groups are balanced with gender, age, intelligence quotient (IQ) score, years of studies, and three cognitive tests' results (WAIS-III, TMT, CVLT) 5 . They are free exchanges carried out in a medical setting where the psychologist is not personally involved -her main action is to maintain the exchange. Preliminary experiments showed that we could distinguish the two groups with relatively high accuracy with the clinician's data. We thus removed clinician's speech turns to reduce this impact and only focused on patients' factors. Further studies are needed to decide how to take into account the entire interaction.", |
| "cite_spans": [ |
| { |
| "start": 111, |
| "end": 134, |
| "text": "(Rebuschi et al., 2014;", |
| "ref_id": "BIBREF37" |
| }, |
| { |
| "start": 135, |
| "end": 156, |
| "text": "Amblard et al., 2015)", |
| "ref_id": "BIBREF2" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Classification: We compare several classification algorithms: Support Vector Machines (SVM) (Cortes and Vapnik, 1995) , Logistic Regression (LR), Random Forest (RF), Perceptron (Perc), and Naive Bayes (NB), without and with feature selection based on importance weight, all implemented in Scikit-Learn (Pedregosa et al., 2011) . Hyperparameters are:", |
| "cite_spans": [ |
| { |
| "start": 92, |
| "end": 117, |
| "text": "(Cortes and Vapnik, 1995)", |
| "ref_id": "BIBREF9" |
| }, |
| { |
| "start": 302, |
| "end": 326, |
| "text": "(Pedregosa et al., 2011)", |
| "ref_id": "BIBREF33" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "\u2022 Naive Bayes: smoothing \u03b1 \u2208 V = {0.001, 0.005, 0.01, 0.1, 0.5, 1, 5, 10, 100}; \u2022 Logistic Regression: L 2 and regularization C \u2208 V ;", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "\u2022 SVM with linear kernel: L 2 and regularization C \u2208 V \u222a {1000};", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "\u2022 Perceptron: L 2 and \u03b1 \u2208 V ;", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "\u2022 Random Forest: max_depth \u2208 {2, N one};", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Thresholds for feature selection are the range of 10 values equally distributed from 1e \u2212 5 to the weight of the 50 th most important feature (thus allowing to keep at least 50 features), plus the mean and median of the weights.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Since our dataset is minimal, we use nested cross-validation to assess the performance of our system: tune hyper-parameters on K \u2212 1 folds and then evaluate on the left-out fold, repeating the whole process M times (K = M = 5). We report average accuracy over the M out folds. Best hyper-parameters values and algorithms are given in Appendix A.2.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Experimental setting", |
| "sec_num": "4" |
| }, |
| { |
| "text": "Lexical features: We compare different representations for Full and Indiv. settings -the most similar to long narrative texts or short Twitter messages. As in previous work, we found that lexical information is very effective (Table 2: bow and ngram) with at best 93.66% in accuracy. However, analysis from previous studies suggested a potential issue: Mitchell et al. (2015) reported that healthrelated lexicon is more represented in the tweets dataset, and Howes et al. (2012b) that the most predictive unigrams are about conditions, treatment and, medication. We investigate our data using Spearman correlation 6 to rank lexical features and find similar results: terms linked to the condition are in top ranks for schizophrenia ( are correlated with controls. This finding is due to the nature of our data: patients talk about their disease with a clinician, and controls talk more about their everyday life. These features perform well because they reflect a lexical bias in data collection. However, the models will not be usable in the wild. Dialogue and discourse: Figure 2 presents results on selected subsets of non-or less-lexicalized features for the five splits of our data. Horizontal lines correspond to the majority vote baselines.", |
| "cite_spans": [ |
| { |
| "start": 353, |
| "end": 375, |
| "text": "Mitchell et al. (2015)", |
| "ref_id": "BIBREF28" |
| }, |
| { |
| "start": 459, |
| "end": 479, |
| "text": "Howes et al. (2012b)", |
| "ref_id": "BIBREF21" |
| }, |
| { |
| "start": 732, |
| "end": 733, |
| "text": "(", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 1073, |
| "end": 1081, |
| "text": "Figure 2", |
| "ref_id": "FIGREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Concerning dialogue features, OCR gives poor results mostly behind the baseline, while BC is above with 74.48% (Full). Moreover, combining with BC to another feature set almost consistently allows improvements (not the case with OCR). These features are good indicators, contrary to what was reported in (Howes et al., 2012b) . Note that we directly use the tokens as features rather than the proportion of BC per word, which allows more refined analysis. The most informative features for controls are phatic expressions (ah, ok, humhum, vraiment [really] POS tags and syntax: Sequences of POS tags (2-POS and 3-POS) and of treelet (2-treelet and 3treelet) are fully non-lexicalized features. They capture some internal structure of the interaction. We obtain our best scores with the longest sequences (3-POS, 72.55% acc., 74.34% F 1 ). These scores are higher than the ones reported by Kayi et al. (2017) on tweets (69.20% F 1 ) or essays (69.76% F 1 ) with simple POS tags and a lot more documents, and are very close to Allende-Cid et al. (2019) with meta-POS (75.1% in F 1 ): this confirms the predictive power of POS for the task.", |
| "cite_spans": [ |
| { |
| "start": 304, |
| "end": 325, |
| "text": "(Howes et al., 2012b)", |
| "ref_id": "BIBREF21" |
| }, |
| { |
| "start": 522, |
| "end": 556, |
| "text": "(ah, ok, humhum, vraiment [really]", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "5" |
| }, |
| { |
| "text": "We found that patients with schizophrenia used more verbs than controls (e.g., 2-POS such as VERB-ADP 7 , 3-POS such as PRON-AUX-VERB), and, as in (Kayi et al., 2017), a higher proportion of adverbs. Precisely, we observe that the usage of adverbs of time (parfois [sometimes], plus maintenant [not anymore], quasiment jamais [almost never]), of place (ici d\u00e9j\u00e0 [here already]) and of frequency and manner (beaucoup plus [much more], beaucoup mieux [much better]) is higher than that of controls -this is possibly linked to the exchange about their (current) heath condition. On the other hand, controls employ a higher portion of linking adverbs (enfin [finally], donc [so], quand m\u00eame [anyway]).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Syntactic features confirm these observations, the most predictive being verbal structures, followed by adverbial modifiers (advmod, advcl) 8 . This goes along with (Kayi et al., 2017) , in which the top parse tag is advmod, and confirms clinician's descriptions on the use of less complex syntactic structures for patients with schizophrenia. Controls tend to use more complicated syntactic structures, such as those with SCONJ (subordinating conjunction) and CCONJ (coordinating conjunction), confirmed by our analysis of discourse connectives.", |
| "cite_spans": [ |
| { |
| "start": 165, |
| "end": 184, |
| "text": "(Kayi et al., 2017)", |
| "ref_id": "BIBREF23" |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Context window size: Our experiments were also designed to test the impact of the context when dealing with dialogues. Figure 2 demonstrates that, in general, the larger the window, the better the scores. Individual speech turns are too small and contain no context. However, using the whole conversation most often leads to a drop in performance compared to our largest window (512 words) due to data sparsity, as we can observe for connective, n-POS and n-treelet. OCR and Backchannels do not follow this trend, meaning that they are probably less sparse.", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 119, |
| "end": 127, |
| "text": "Figure 2", |
| "ref_id": "FIGREF1" |
| } |
| ], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "5" |
| }, |
| { |
| "text": "These experiments demonstrate that using the block of conversation is relevant -the models find enough information to make accurate classification -, while allowing to increase the number of classification instances artificially.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "5" |
| }, |
| { |
| "text": "Best algorithm: Among the 5 classifiers, NB generally performs well when dealing with word counts (in Full and Indiv.), while SVM and LR are generally better in other cases. More precisely, SVM performs better when the context window is relatively large, and the data sparsity is more pronounced (Full). At the same time, LR is better at dealing with small to medium-sized contexts (Indiv. and W-n settings). Detailed information is in the supplementary material.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Results", |
| "sec_num": "5" |
| }, |
| { |
| "text": "We used conversations involving patients with schizophrenia in order to learn about language features associated with the disease. We compared various settings to represent dialogues and several representations to deal with data scarcity and lexical bias. Our experiments replicate performances as high as previous studies in English. Further experiments will be designed to take into account the entire interaction, probably with neural networks. We would also like to investigate the effect of adversarial loss in mitigating the bias within a neural model.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "6" |
| }, |
| { |
| "text": "We hope that this paper will remind us of the importance of looking for bias in data and exploring higher-level, less language-dependent information to produce robust systems and draw more general conclusions on conversational data. ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Conclusion", |
| "sec_num": "6" |
| }, |
| { |
| "text": "A.1 OCR and backchannel word list", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "A Appendices", |
| "sec_num": null |
| }, |
| { |
| "text": "In order to improve reproducibility, we give the full list of tokens used for OCR (Table 3) and backchannel (Table 5) , as well as their corresponding translation in English (Table 4, Table 6 ). They were obtained by translating the list given by the authors of we contacted, and by adding a few additional terms specific to French. A.2 Best scores and corresp. settings Table 7 : Best scores (averaged accuracy Acc.), best algorithms (Algo), corresponding hyper-parameters (Hyperparams.) and threshold (Thres.) for full documents (Full), individual speech turns (Indiv.) and Window size of 512 tokens (W-512).", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 82, |
| "end": 91, |
| "text": "(Table 3)", |
| "ref_id": "TABREF8" |
| }, |
| { |
| "start": 108, |
| "end": 117, |
| "text": "(Table 5)", |
| "ref_id": "TABREF10" |
| }, |
| { |
| "start": 174, |
| "end": 191, |
| "text": "(Table 4, Table 6", |
| "ref_id": "TABREF9" |
| }, |
| { |
| "start": 371, |
| "end": 378, |
| "text": "Table 7", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "A Appendices", |
| "sec_num": null |
| }, |
| { |
| "text": "Our code is on: https://github.com/ chuyuanli/non-lexical-markers-scz-conv.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "POS tagging is a process of marking up a word in a text to a particular part of speech. Allende-Cid et al. (2019) tested two types of POS tags: a general one called meta-POS (12 labels) and a precise one POS (160 labels). Both allow performance higher than chance.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "WAIS-III: Wechsler Adult Intelligence Scale (WAIS) is an IQ test designed to measure intelligence and cognitive ability in adults and older adolescents. Trail Making Test (TMT) is a widely used test to assess executive abilities in patients. California Verbal Learning Test (CVLT) measures episodic verbal learning and memory.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "p-value< 0.05, coefficient |\u03c1| > 0.3", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "ADP stands for adposition and it covers preposition and postposition.8 advmod is a (non-clausal) adverb or adverbial phrase; advcl is an adverbial clause modifier. They serve to modify a verb or other predicate.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [ |
| { |
| "text": "The authors thank the anonymous reviewers for their insightful comments and suggestions. This work was supported by the PIA project \"Lorraine Universit\u00e9 d'Excellence\", ANR-15-IDEX-04-LUE, as well as the CPER LCHN (Contrat de Plan \u00c9tat-R\u00e9gion -Langues, Connaissances et Humanit\u00e9s Num\u00e9riques). We would like to thank the Centre Hospitalier Le Vinartier for having contributed decisively to the implementation of the experimentation. Experiments presented in this paper were carried out in secured nodes on the Grid'5000 testbed. We would like to thank the Grid'5000 community (see https://www.grid5000.fr).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Acknowledgements", |
| "sec_num": null |
| } |
| ], |
| "bib_entries": { |
| "BIBREF0": { |
| "ref_id": "b0", |
| "title": "A machine learning approach for the automatic classification of schizophrenic discourse", |
| "authors": [ |
| { |
| "first": "Hector", |
| "middle": [], |
| "last": "Allende-Cid", |
| "suffix": "" |
| }, |
| { |
| "first": "Juan", |
| "middle": [], |
| "last": "Zamora", |
| "suffix": "" |
| }, |
| { |
| "first": "Pedro", |
| "middle": [], |
| "last": "Alfaron-Faccio", |
| "suffix": "" |
| }, |
| { |
| "first": "Mar\u00eda", |
| "middle": [], |
| "last": "Alonso", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "IEEE Access", |
| "volume": "", |
| "issue": "", |
| "pages": "45544--45554", |
| "other_ids": { |
| "DOI": [ |
| "10.1109/ACCESS.2019.2908620" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Hector Allende-Cid, Juan Zamora, Pedro Alfaron- Faccio, and Mar\u00eda Alonso. 2019. A machine learning approach for the automatic classification of schizophrenic discourse. IEEE Access, pages 45544-45554.", |
| "links": null |
| }, |
| "BIBREF1": { |
| "ref_id": "b1", |
| "title": "Investigation par m\u00e9thodes d'apprentissage des sp\u00e9cificit\u00e9s langagi\u00e8res propres aux personnes avec schizophr\u00e9nie (investigating learning methods applied to language specificity of persons with schizophrenia)", |
| "authors": [ |
| { |
| "first": "Maxime", |
| "middle": [], |
| "last": "Amblard", |
| "suffix": "" |
| }, |
| { |
| "first": "Chlo\u00e9", |
| "middle": [], |
| "last": "Braud", |
| "suffix": "" |
| }, |
| { |
| "first": "Chuyuan", |
| "middle": [], |
| "last": "Li", |
| "suffix": "" |
| }, |
| { |
| "first": "Caroline", |
| "middle": [], |
| "last": "Demily", |
| "suffix": "" |
| }, |
| { |
| "first": "Nicolas", |
| "middle": [], |
| "last": "Franck", |
| "suffix": "" |
| }, |
| { |
| "first": "Michel", |
| "middle": [], |
| "last": "Musiol", |
| "suffix": "" |
| } |
| ], |
| "year": 2020, |
| "venue": "Actes de la 6e conf\u00e9rence conjointe Journ\u00e9es d'\u00c9tudes sur la Parole (JEP, 33e \u00e9dition)", |
| "volume": "2", |
| "issue": "", |
| "pages": "12--26", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Maxime Amblard, Chlo\u00e9 Braud, Chuyuan Li, Caro- line Demily, Nicolas Franck, and Michel Musiol. 2020. Investigation par m\u00e9thodes d'apprentissage des sp\u00e9cificit\u00e9s langagi\u00e8res propres aux personnes avec schizophr\u00e9nie (investigating learning methods applied to language specificity of persons with schizophrenia). In Actes de la 6e conf\u00e9rence con- jointe Journ\u00e9es d'\u00c9tudes sur la Parole (JEP, 33e \u00e9dition), Traitement Automatique des Langues Na- turelles (TALN, 27e \u00e9dition), Rencontre des \u00c9tudi- ants Chercheurs en Informatique pour le Traitement Automatique des Langues (R\u00c9CITAL, 22e \u00e9dition). Volume 2: Traitement Automatique des Langues Na- turelles, pages 12-26.", |
| "links": null |
| }, |
| "BIBREF2": { |
| "ref_id": "b2", |
| "title": "Analyse lexicale outill\u00e9e de la parole transcrite de patients schizophr\u00e8nes", |
| "authors": [ |
| { |
| "first": "Maxime", |
| "middle": [], |
| "last": "Amblard", |
| "suffix": "" |
| }, |
| { |
| "first": "Kar\u00ebn", |
| "middle": [], |
| "last": "Fort", |
| "suffix": "" |
| }, |
| { |
| "first": "Caroline", |
| "middle": [], |
| "last": "Demily", |
| "suffix": "" |
| } |
| ], |
| "year": 2015, |
| "venue": "Traitement Automatique des Langues", |
| "volume": "55", |
| "issue": "3", |
| "pages": "91--115", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Maxime Amblard, Kar\u00ebn Fort, Caroline Demily, Nico- las Franck, and Michel Musiol. 2015. Anal- yse lexicale outill\u00e9e de la parole transcrite de pa- tients schizophr\u00e8nes. Traitement Automatique des Langues, 55(3):91 -115.", |
| "links": null |
| }, |
| "BIBREF3": { |
| "ref_id": "b3", |
| "title": "DSM-5-Manuel diagnostique et statistique des troubles mentaux", |
| "authors": [], |
| "year": 2015, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "American Psychiatric Association APA. 2015. DSM-5- Manuel diagnostique et statistique des troubles men- taux. Elsevier Masson.", |
| "links": null |
| }, |
| "BIBREF4": { |
| "ref_id": "b4", |
| "title": "Multitask learning for mental health conditions with limited social media data", |
| "authors": [ |
| { |
| "first": "Adrian", |
| "middle": [], |
| "last": "Benton", |
| "suffix": "" |
| }, |
| { |
| "first": "Margaret", |
| "middle": [], |
| "last": "Mitchell", |
| "suffix": "" |
| }, |
| { |
| "first": "Dirk", |
| "middle": [], |
| "last": "Hovy", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics", |
| "volume": "1", |
| "issue": "", |
| "pages": "152--162", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Adrian Benton, Margaret Mitchell, and Dirk Hovy. 2017. Multitask learning for mental health condi- tions with limited social media data. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Vol- ume 1, Long Papers, pages 152-162.", |
| "links": null |
| }, |
| "BIBREF5": { |
| "ref_id": "b5", |
| "title": "A collaborative approach to identifying social media markers of schizophrenia by employing machine learning and clinical appraisals", |
| "authors": [ |
| { |
| "first": "L", |
| "middle": [], |
| "last": "Michael", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Birnbaum", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Sindhu Kiranmai Ernala", |
| "suffix": "" |
| }, |
| { |
| "first": "F", |
| "middle": [], |
| "last": "Asra", |
| "suffix": "" |
| }, |
| { |
| "first": "Munmun", |
| "middle": [ |
| "De" |
| ], |
| "last": "Rizvi", |
| "suffix": "" |
| }, |
| { |
| "first": "John M", |
| "middle": [], |
| "last": "Choudhury", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Kane", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "J Med Internet Res", |
| "volume": "19", |
| "issue": "8", |
| "pages": "", |
| "other_ids": { |
| "DOI": [ |
| "10.2196/jmir.7956" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Michael L Birnbaum, Sindhu Kiranmai Ernala, Asra F Rizvi, Munmun De Choudhury, and John M Kane. 2017. A collaborative approach to identifying so- cial media markers of schizophrenia by employing machine learning and clinical appraisals. J Med In- ternet Res, 19(8):e289.", |
| "links": null |
| }, |
| "BIBREF6": { |
| "ref_id": "b6", |
| "title": "Latent dirichlet allocation", |
| "authors": [ |
| { |
| "first": "M", |
| "middle": [], |
| "last": "David", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Blei", |
| "suffix": "" |
| }, |
| { |
| "first": "Y", |
| "middle": [], |
| "last": "Andrew", |
| "suffix": "" |
| }, |
| { |
| "first": "Michael I Jordan", |
| "middle": [], |
| "last": "Ng", |
| "suffix": "" |
| } |
| ], |
| "year": 2003, |
| "venue": "Journal of machine Learning research", |
| "volume": "3", |
| "issue": "", |
| "pages": "993--1022", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "David M Blei, Andrew Y Ng, and Michael I Jordan. 2003. Latent dirichlet allocation. Journal of ma- chine Learning research, 3(Jan):993-1022.", |
| "links": null |
| }, |
| "BIBREF7": { |
| "ref_id": "b7", |
| "title": "Classbased n-gram models of natural language", |
| "authors": [ |
| { |
| "first": "Vincent J Della", |
| "middle": [], |
| "last": "Peter F Brown", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Pietra", |
| "suffix": "" |
| }, |
| { |
| "first": "V", |
| "middle": [], |
| "last": "Peter", |
| "suffix": "" |
| }, |
| { |
| "first": "Jennifer", |
| "middle": [ |
| "C" |
| ], |
| "last": "Desouza", |
| "suffix": "" |
| }, |
| { |
| "first": "Robert L", |
| "middle": [], |
| "last": "Lai", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Mercer", |
| "suffix": "" |
| } |
| ], |
| "year": 1992, |
| "venue": "Computational linguistics", |
| "volume": "18", |
| "issue": "4", |
| "pages": "467--480", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Peter F Brown, Vincent J Della Pietra, Peter V Desouza, Jennifer C Lai, and Robert L Mercer. 1992. Class- based n-gram models of natural language. Compu- tational linguistics, 18(4):467-480.", |
| "links": null |
| }, |
| "BIBREF8": { |
| "ref_id": "b8", |
| "title": "A linguist looks at \"schizophrenic\" language", |
| "authors": [ |
| { |
| "first": "Elaine", |
| "middle": [], |
| "last": "Chaika", |
| "suffix": "" |
| } |
| ], |
| "year": 1974, |
| "venue": "Brain and language", |
| "volume": "1", |
| "issue": "3", |
| "pages": "257--276", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Elaine Chaika. 1974. A linguist looks at \"schizophrenic\" language. Brain and language, 1(3):257-276.", |
| "links": null |
| }, |
| "BIBREF9": { |
| "ref_id": "b9", |
| "title": "Supportvector networks", |
| "authors": [ |
| { |
| "first": "Corinna", |
| "middle": [], |
| "last": "Cortes", |
| "suffix": "" |
| }, |
| { |
| "first": "Vladimir", |
| "middle": [], |
| "last": "Vapnik", |
| "suffix": "" |
| } |
| ], |
| "year": 1995, |
| "venue": "Machine learning", |
| "volume": "20", |
| "issue": "3", |
| "pages": "273--297", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Corinna Cortes and Vladimir Vapnik. 1995. Support- vector networks. Machine learning, 20(3):273-297.", |
| "links": null |
| }, |
| "BIBREF10": { |
| "ref_id": "b10", |
| "title": "Communication disturbances in schizophrenia and mania", |
| "authors": [ |
| { |
| "first": "M", |
| "middle": [], |
| "last": "Nancy", |
| "suffix": "" |
| }, |
| { |
| "first": "Maddalena", |
| "middle": [], |
| "last": "Docherty", |
| "suffix": "" |
| }, |
| { |
| "first": "Nancy", |
| "middle": [ |
| "C" |
| ], |
| "last": "Derosa", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Andreasen", |
| "suffix": "" |
| } |
| ], |
| "year": 1996, |
| "venue": "Archives of General Psychiatry", |
| "volume": "53", |
| "issue": "4", |
| "pages": "358--364", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Nancy M Docherty, Maddalena DeRosa, and Nancy C Andreasen. 1996. Communication disturbances in schizophrenia and mania. Archives of General Psy- chiatry, 53(4):358-364.", |
| "links": null |
| }, |
| "BIBREF11": { |
| "ref_id": "b11", |
| "title": "Quantifying incoherence in speech: an automated methodology and novel application to schizophrenia", |
| "authors": [ |
| { |
| "first": "Brita", |
| "middle": [], |
| "last": "Elvev\u00e5g", |
| "suffix": "" |
| }, |
| { |
| "first": "W", |
| "middle": [], |
| "last": "Peter", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Foltz", |
| "suffix": "" |
| }, |
| { |
| "first": "Terry", |
| "middle": [ |
| "E" |
| ], |
| "last": "Daniel R Weinberger", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Goldberg", |
| "suffix": "" |
| } |
| ], |
| "year": 2007, |
| "venue": "Schizophrenia research", |
| "volume": "93", |
| "issue": "1-3", |
| "pages": "304--316", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Brita Elvev\u00e5g, Peter W Foltz, Daniel R Weinberger, and Terry E Goldberg. 2007. Quantifying inco- herence in speech: an automated methodology and novel application to schizophrenia. Schizophrenia research, 93(1-3):304-316.", |
| "links": null |
| }, |
| "BIBREF12": { |
| "ref_id": "b12", |
| "title": "Linguistic features identify alzheimer's disease in narrative speech", |
| "authors": [ |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Kathleen", |
| "suffix": "" |
| }, |
| { |
| "first": "Jed", |
| "middle": [ |
| "A" |
| ], |
| "last": "Fraser", |
| "suffix": "" |
| }, |
| { |
| "first": "Frank", |
| "middle": [], |
| "last": "Meltzer", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Rudzicz", |
| "suffix": "" |
| } |
| ], |
| "year": 2016, |
| "venue": "Journal of Alzheimer's Disease", |
| "volume": "49", |
| "issue": "2", |
| "pages": "407--422", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Kathleen C Fraser, Jed A Meltzer, and Frank Rudzicz. 2016. Linguistic features identify alzheimer's dis- ease in narrative speech. Journal of Alzheimer's Dis- ease, 49(2):407-422.", |
| "links": null |
| }, |
| "BIBREF13": { |
| "ref_id": "b13", |
| "title": "The diagnosis of schizophrenia by language analysis", |
| "authors": [ |
| { |
| "first": "I", |
| "middle": [], |
| "last": "William", |
| "suffix": "" |
| }, |
| { |
| "first": "Kathleen", |
| "middle": [ |
| "M" |
| ], |
| "last": "Fraser", |
| "suffix": "" |
| }, |
| { |
| "first": "Philip", |
| "middle": [], |
| "last": "King", |
| "suffix": "" |
| }, |
| { |
| "first": "Robert", |
| "middle": [ |
| "E" |
| ], |
| "last": "Thomas", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Kendell", |
| "suffix": "" |
| } |
| ], |
| "year": 1986, |
| "venue": "The British Journal of Psychiatry", |
| "volume": "148", |
| "issue": "3", |
| "pages": "275--278", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "William I Fraser, Kathleen M King, Philip Thomas, and Robert E Kendell. 1986. The diagnosis of schizophrenia by language analysis. The British Journal of Psychiatry, 148(3):275-278.", |
| "links": null |
| }, |
| "BIBREF14": { |
| "ref_id": "b14", |
| "title": "Detecting language impairments in autism: A computational analysis of semi-structured conversations with vector semantics", |
| "authors": [ |
| { |
| "first": "Adam", |
| "middle": [], |
| "last": "Goodkind", |
| "suffix": "" |
| }, |
| { |
| "first": "Michelle", |
| "middle": [], |
| "last": "Lee", |
| "suffix": "" |
| }, |
| { |
| "first": "E", |
| "middle": [], |
| "last": "Gary", |
| "suffix": "" |
| }, |
| { |
| "first": "Molly", |
| "middle": [], |
| "last": "Martin", |
| "suffix": "" |
| }, |
| { |
| "first": "Klinton", |
| "middle": [], |
| "last": "Losh", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Bicknell", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Proceedings of the Society for Computation in Linguistics", |
| "volume": "1", |
| "issue": "1", |
| "pages": "12--22", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Adam Goodkind, Michelle Lee, Gary E Martin, Molly Losh, and Klinton Bicknell. 2018. Detecting lan- guage impairments in autism: A computational anal- ysis of semi-structured conversations with vector se- mantics. Proceedings of the Society for Computa- tion in Linguistics, 1(1):12-22.", |
| "links": null |
| }, |
| "BIBREF15": { |
| "ref_id": "b15", |
| "title": "What twitter profile and posted images reveal about depression and anxiety", |
| "authors": [ |
| { |
| "first": "Chandra", |
| "middle": [], |
| "last": "Sharath", |
| "suffix": "" |
| }, |
| { |
| "first": "Daniel", |
| "middle": [], |
| "last": "Guntuku", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Preotiuc-Pietro", |
| "suffix": "" |
| }, |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Johannes", |
| "suffix": "" |
| }, |
| { |
| "first": "Lyle", |
| "middle": [ |
| "H" |
| ], |
| "last": "Eichstaedt", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Ungar", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Proceedings of the International AAAI Conference on Web and Social Media", |
| "volume": "13", |
| "issue": "", |
| "pages": "236--246", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Sharath Chandra Guntuku, Daniel Preotiuc-Pietro, Jo- hannes C Eichstaedt, and Lyle H Ungar. 2019. What twitter profile and posted images reveal about de- pression and anxiety. In Proceedings of the Interna- tional AAAI Conference on Web and Social Media, volume 13, pages 236-246.", |
| "links": null |
| }, |
| "BIBREF16": { |
| "ref_id": "b16", |
| "title": "An analysis of grammatical deviance occurring in spontaneous schizophrenic speech", |
| "authors": [ |
| { |
| "first": "E", |
| "middle": [], |
| "last": "Ralph", |
| "suffix": "" |
| }, |
| { |
| "first": "William", |
| "middle": [], |
| "last": "Hoffman", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Sledge", |
| "suffix": "" |
| } |
| ], |
| "year": 1988, |
| "venue": "Journal of neurolinguistics", |
| "volume": "3", |
| "issue": "1", |
| "pages": "89--101", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ralph E Hoffman and William Sledge. 1988. An anal- ysis of grammatical deviance occurring in sponta- neous schizophrenic speech. Journal of neurolin- guistics, 3(1):89-101.", |
| "links": null |
| }, |
| "BIBREF17": { |
| "ref_id": "b17", |
| "title": "Lexical differences in autobiographical narratives from schizophrenic patients and healthy controls", |
| "authors": [ |
| { |
| "first": "Kai", |
| "middle": [], |
| "last": "Hong", |
| "suffix": "" |
| }, |
| { |
| "first": "G", |
| "middle": [], |
| "last": "Christian", |
| "suffix": "" |
| }, |
| { |
| "first": "Mary", |
| "middle": [ |
| "E" |
| ], |
| "last": "Kohler", |
| "suffix": "" |
| }, |
| { |
| "first": "Amber", |
| "middle": [ |
| "A" |
| ], |
| "last": "March", |
| "suffix": "" |
| }, |
| { |
| "first": "Ani", |
| "middle": [], |
| "last": "Parker", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Nenkova", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning", |
| "volume": "", |
| "issue": "", |
| "pages": "37--47", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Kai Hong, Christian G Kohler, Mary E March, Am- ber A Parker, and Ani Nenkova. 2012. Lexi- cal differences in autobiographical narratives from schizophrenic patients and healthy controls. In Pro- ceedings of the 2012 Joint Conference on Empiri- cal Methods in Natural Language Processing and Computational Natural Language Learning, pages 37-47.", |
| "links": null |
| }, |
| "BIBREF18": { |
| "ref_id": "b18", |
| "title": "Helping the medicine go down: Repair and adherence in patient-clinician dialogues", |
| "authors": [ |
| { |
| "first": "Christine", |
| "middle": [], |
| "last": "Howes", |
| "suffix": "" |
| }, |
| { |
| "first": "Matt", |
| "middle": [], |
| "last": "Purver", |
| "suffix": "" |
| }, |
| { |
| "first": "Rose", |
| "middle": [], |
| "last": "Mccabe", |
| "suffix": "" |
| }, |
| { |
| "first": "G", |
| "middle": [ |
| "T" |
| ], |
| "last": "Patrick", |
| "suffix": "" |
| }, |
| { |
| "first": "Mary", |
| "middle": [], |
| "last": "Healey", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Lavelle", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of SemDial 2012 (SeineDial): The 16th Workshop on the Semantics and Pragmatics of Dialogue", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Christine Howes, Matt Purver, Rose McCabe, Patrick GT Healey, and Mary Lavelle. 2012a. Help- ing the medicine go down: Repair and adherence in patient-clinician dialogues. In Proceedings of SemDial 2012 (SeineDial): The 16th Workshop on the Semantics and Pragmatics of Dialogue, page 155.", |
| "links": null |
| }, |
| "BIBREF19": { |
| "ref_id": "b19", |
| "title": "Using conversation topics for predicting therapy outcomes in schizophrenia", |
| "authors": [ |
| { |
| "first": "Christine", |
| "middle": [], |
| "last": "Howes", |
| "suffix": "" |
| }, |
| { |
| "first": "Matthew", |
| "middle": [], |
| "last": "Purver", |
| "suffix": "" |
| }, |
| { |
| "first": "Rose", |
| "middle": [], |
| "last": "Mccabe", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "Biomedical informatics insights", |
| "volume": "6", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Christine Howes, Matthew Purver, and Rose McCabe. 2013. Using conversation topics for predicting ther- apy outcomes in schizophrenia. Biomedical infor- matics insights, 6:BII-S11661.", |
| "links": null |
| }, |
| "BIBREF20": { |
| "ref_id": "b20", |
| "title": "Linguistic indicators of severity and progress in online text-based therapy for depression", |
| "authors": [ |
| { |
| "first": "Christine", |
| "middle": [], |
| "last": "Howes", |
| "suffix": "" |
| }, |
| { |
| "first": "Matthew", |
| "middle": [], |
| "last": "Purver", |
| "suffix": "" |
| }, |
| { |
| "first": "Rose", |
| "middle": [], |
| "last": "Mccabe", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Christine Howes, Matthew Purver, and Rose McCabe. 2014. Linguistic indicators of severity and progress in online text-based therapy for depression. ACL 2014, page 7.", |
| "links": null |
| }, |
| "BIBREF21": { |
| "ref_id": "b21", |
| "title": "Predicting adherence to treatment for schizophrenia from dialogue transcripts", |
| "authors": [ |
| { |
| "first": "Christine", |
| "middle": [], |
| "last": "Howes", |
| "suffix": "" |
| }, |
| { |
| "first": "Matthew", |
| "middle": [], |
| "last": "Purver", |
| "suffix": "" |
| }, |
| { |
| "first": "Rose", |
| "middle": [], |
| "last": "Mccabe", |
| "suffix": "" |
| }, |
| { |
| "first": "Patrick", |
| "middle": [], |
| "last": "Healey", |
| "suffix": "" |
| }, |
| { |
| "first": "Mary", |
| "middle": [], |
| "last": "Lavelle", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue", |
| "volume": "", |
| "issue": "", |
| "pages": "79--83", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Christine Howes, Matthew Purver, Rose McCabe, Patrick Healey, and Mary Lavelle. 2012b. Predict- ing adherence to treatment for schizophrenia from dialogue transcripts. In Proceedings of the 13th An- nual Meeting of the Special Interest Group on Dis- course and Dialogue, pages 79-83.", |
| "links": null |
| }, |
| "BIBREF22": { |
| "ref_id": "b22", |
| "title": "Cross-lingual syntactic variation over age and gender", |
| "authors": [ |
| { |
| "first": "Anders", |
| "middle": [], |
| "last": "Johannsen", |
| "suffix": "" |
| }, |
| { |
| "first": "Dirk", |
| "middle": [], |
| "last": "Hovy", |
| "suffix": "" |
| }, |
| { |
| "first": "Anders", |
| "middle": [], |
| "last": "S\u00f8gaard", |
| "suffix": "" |
| } |
| ], |
| "year": 2015, |
| "venue": "Proceedings of the nineteenth conference on computational natural language learning", |
| "volume": "", |
| "issue": "", |
| "pages": "103--112", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Anders Johannsen, Dirk Hovy, and Anders S\u00f8gaard. 2015. Cross-lingual syntactic variation over age and gender. In Proceedings of the nineteenth conference on computational natural language learning, pages 103-112.", |
| "links": null |
| }, |
| "BIBREF23": { |
| "ref_id": "b23", |
| "title": "Predictive linguistic features of schizophrenia", |
| "authors": [ |
| { |
| "first": "Efsun", |
| "middle": [], |
| "last": "Sarioglu Kayi", |
| "suffix": "" |
| }, |
| { |
| "first": "Mona", |
| "middle": [], |
| "last": "Diab", |
| "suffix": "" |
| }, |
| { |
| "first": "Luca", |
| "middle": [], |
| "last": "Pauselli", |
| "suffix": "" |
| }, |
| { |
| "first": "Michael", |
| "middle": [], |
| "last": "Compton", |
| "suffix": "" |
| }, |
| { |
| "first": "Glen", |
| "middle": [], |
| "last": "Coppersmith", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (* SEM 2017)", |
| "volume": "", |
| "issue": "", |
| "pages": "241--250", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Efsun Sarioglu Kayi, Mona Diab, Luca Pauselli, Michael Compton, and Glen Coppersmith. 2017. Predictive linguistic features of schizophrenia. In Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (* SEM 2017), pages 241-250.", |
| "links": null |
| }, |
| "BIBREF24": { |
| "ref_id": "b24", |
| "title": "Early linguistic markers of trauma-specific processing predict post-trauma adjustment", |
| "authors": [ |
| { |
| "first": "Birgit", |
| "middle": [], |
| "last": "Kleim", |
| "suffix": "" |
| }, |
| { |
| "first": "Andrea", |
| "middle": [ |
| "B" |
| ], |
| "last": "Horn", |
| "suffix": "" |
| }, |
| { |
| "first": "Rainer", |
| "middle": [], |
| "last": "Kraehenmann", |
| "suffix": "" |
| }, |
| { |
| "first": "Matthias", |
| "middle": [ |
| "R" |
| ], |
| "last": "Mehl", |
| "suffix": "" |
| }, |
| { |
| "first": "Anke", |
| "middle": [], |
| "last": "Ehlers", |
| "suffix": "" |
| } |
| ], |
| "year": 2018, |
| "venue": "Frontiers in Psychiatry", |
| "volume": "9", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Birgit Kleim, Andrea B Horn, Rainer Kraehenmann, Matthias R Mehl, and Anke Ehlers. 2018. Early linguistic markers of trauma-specific processing pre- dict post-trauma adjustment. Frontiers in Psychia- try, 9:645.", |
| "links": null |
| }, |
| "BIBREF25": { |
| "ref_id": "b25", |
| "title": "Language in schizophrenia part 1: an introduction", |
| "authors": [ |
| { |
| "first": "", |
| "middle": [], |
| "last": "Gina R Kuperberg", |
| "suffix": "" |
| } |
| ], |
| "year": 2010, |
| "venue": "Language and linguistics compass", |
| "volume": "4", |
| "issue": "8", |
| "pages": "576--589", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Gina R Kuperberg. 2010. Language in schizophrenia part 1: an introduction. Language and linguistics compass, 4(8):576-589.", |
| "links": null |
| }, |
| "BIBREF26": { |
| "ref_id": "b26", |
| "title": "Quantitative assessment of the frequency of normal associations in the utterances of schizophrenia patients and healthy controls", |
| "authors": [ |
| { |
| "first": "A", |
| "middle": [], |
| "last": "Brendan", |
| "suffix": "" |
| }, |
| { |
| "first": "Theo", |
| "middle": [ |
| "C" |
| ], |
| "last": "Maher", |
| "suffix": "" |
| }, |
| { |
| "first": "Jakob", |
| "middle": [], |
| "last": "Manschreck", |
| "suffix": "" |
| }, |
| { |
| "first": "Steven", |
| "middle": [], |
| "last": "Linnet", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Candela", |
| "suffix": "" |
| } |
| ], |
| "year": 2005, |
| "venue": "Schizophrenia Research", |
| "volume": "78", |
| "issue": "2-3", |
| "pages": "219--224", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Brendan A Maher, Theo C Manschreck, Jakob Linnet, and Steven Candela. 2005. Quantitative assessment of the frequency of normal associations in the utter- ances of schizophrenia patients and healthy controls. Schizophrenia Research, 78(2-3):219-224.", |
| "links": null |
| }, |
| "BIBREF27": { |
| "ref_id": "b27", |
| "title": "Reduced primacy and related features in schizophrenia", |
| "authors": [ |
| { |
| "first": "C", |
| "middle": [], |
| "last": "Theo", |
| "suffix": "" |
| }, |
| { |
| "first": "Brendan", |
| "middle": [ |
| "A" |
| ], |
| "last": "Manschreck", |
| "suffix": "" |
| }, |
| { |
| "first": "Jill", |
| "middle": [ |
| "E" |
| ], |
| "last": "Maher", |
| "suffix": "" |
| }, |
| { |
| "first": "Jonathan", |
| "middle": [], |
| "last": "Rosenthal", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Berner", |
| "suffix": "" |
| } |
| ], |
| "year": 1991, |
| "venue": "Schizophrenia Research", |
| "volume": "5", |
| "issue": "1", |
| "pages": "35--41", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Theo C Manschreck, Brendan A Maher, Jill E Rosen- thal, and Jonathan Berner. 1991. Reduced primacy and related features in schizophrenia. Schizophrenia Research, 5(1):35-41.", |
| "links": null |
| }, |
| "BIBREF28": { |
| "ref_id": "b28", |
| "title": "Quantifying the language of schizophrenia in social media", |
| "authors": [ |
| { |
| "first": "Margaret", |
| "middle": [], |
| "last": "Mitchell", |
| "suffix": "" |
| }, |
| { |
| "first": "Kristy", |
| "middle": [], |
| "last": "Hollingshead", |
| "suffix": "" |
| }, |
| { |
| "first": "Glen", |
| "middle": [], |
| "last": "Coppersmith", |
| "suffix": "" |
| } |
| ], |
| "year": 2015, |
| "venue": "Proceedings of the 2nd workshop on Computational linguistics and clinical psychology: From linguistic signal to clinical reality", |
| "volume": "", |
| "issue": "", |
| "pages": "11--20", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Margaret Mitchell, Kristy Hollingshead, and Glen Coppersmith. 2015. Quantifying the language of schizophrenia in social media. In Proceedings of the 2nd workshop on Computational linguistics and clinical psychology: From linguistic signal to clini- cal reality, pages 11-20.", |
| "links": null |
| }, |
| "BIBREF29": { |
| "ref_id": "b29", |
| "title": "Language analysis in schizophrenia: Diagnostic implications", |
| "authors": [ |
| { |
| "first": "D", |
| "middle": [], |
| "last": "Rodney", |
| "suffix": "" |
| }, |
| { |
| "first": "John Cl", |
| "middle": [], |
| "last": "Morice", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Ingram", |
| "suffix": "" |
| } |
| ], |
| "year": 1982, |
| "venue": "Australian and New Zealand Journal of Psychiatry", |
| "volume": "16", |
| "issue": "2", |
| "pages": "11--21", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Rodney D Morice and John CL Ingram. 1982. Lan- guage analysis in schizophrenia: Diagnostic implica- tions. Australian and New Zealand Journal of Psy- chiatry, 16(2):11-21.", |
| "links": null |
| }, |
| "BIBREF30": { |
| "ref_id": "b30", |
| "title": "El\u00e9ments de psychopathologie cognitive: le discours schizophr\u00e8ne", |
| "authors": [ |
| { |
| "first": "Michel", |
| "middle": [], |
| "last": "Musiol", |
| "suffix": "" |
| }, |
| { |
| "first": "Alain", |
| "middle": [], |
| "last": "Trognon", |
| "suffix": "" |
| } |
| ], |
| "year": 2000, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Michel Musiol and Alain Trognon. 2000. El\u00e9- ments de psychopathologie cognitive: le discours schizophr\u00e8ne. A. Colin.", |
| "links": null |
| }, |
| "BIBREF31": { |
| "ref_id": "b31", |
| "title": "Learning predictive linguistic features for alzheimer's disease and related dementias using verbal utterances", |
| "authors": [ |
| { |
| "first": "Jojo", |
| "middle": [], |
| "last": "Sylvester Olubolu Orimaye", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Sze-Meng", |
| "suffix": "" |
| }, |
| { |
| "first": "Karen", |
| "middle": [ |
| "Jennifer" |
| ], |
| "last": "Wong", |
| "suffix": "" |
| }, |
| { |
| "first": "", |
| "middle": [], |
| "last": "Golden", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "Proceedings of the workshop on computational linguistics and clinical psychology: From linguistic signal to clinical reality", |
| "volume": "", |
| "issue": "", |
| "pages": "78--87", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Sylvester Olubolu Orimaye, Jojo Sze-Meng Wong, and Karen Jennifer Golden. 2014. Learning predictive linguistic features for alzheimer's disease and re- lated dementias using verbal utterances. In Proceed- ings of the workshop on computational linguistics and clinical psychology: From linguistic signal to clinical reality, pages 78-87.", |
| "links": null |
| }, |
| "BIBREF32": { |
| "ref_id": "b32", |
| "title": "Screening twitter users for depression and ptsd with lexical decision lists", |
| "authors": [ |
| { |
| "first": "Ted", |
| "middle": [], |
| "last": "Pedersen", |
| "suffix": "" |
| } |
| ], |
| "year": 2015, |
| "venue": "Proceedings of the 2nd workshop on computational linguistics and clinical psychology: from linguistic signal to clinical reality", |
| "volume": "", |
| "issue": "", |
| "pages": "46--53", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Ted Pedersen. 2015. Screening twitter users for depres- sion and ptsd with lexical decision lists. In Proceed- ings of the 2nd workshop on computational linguis- tics and clinical psychology: from linguistic signal to clinical reality, pages 46-53.", |
| "links": null |
| }, |
| "BIBREF33": { |
| "ref_id": "b33", |
| "title": "Scikit-learn: Machine learning in python. the", |
| "authors": [ |
| { |
| "first": "Fabian", |
| "middle": [], |
| "last": "Pedregosa", |
| "suffix": "" |
| }, |
| { |
| "first": "Ga\u00ebl", |
| "middle": [], |
| "last": "Varoquaux", |
| "suffix": "" |
| }, |
| { |
| "first": "Alexandre", |
| "middle": [], |
| "last": "Gramfort", |
| "suffix": "" |
| }, |
| { |
| "first": "Vincent", |
| "middle": [], |
| "last": "Michel", |
| "suffix": "" |
| }, |
| { |
| "first": "Bertrand", |
| "middle": [], |
| "last": "Thirion", |
| "suffix": "" |
| }, |
| { |
| "first": "Olivier", |
| "middle": [], |
| "last": "Grisel", |
| "suffix": "" |
| }, |
| { |
| "first": "Mathieu", |
| "middle": [], |
| "last": "Blondel", |
| "suffix": "" |
| }, |
| { |
| "first": "Peter", |
| "middle": [], |
| "last": "Prettenhofer", |
| "suffix": "" |
| }, |
| { |
| "first": "Ron", |
| "middle": [], |
| "last": "Weiss", |
| "suffix": "" |
| }, |
| { |
| "first": "Vincent", |
| "middle": [], |
| "last": "Dubourg", |
| "suffix": "" |
| } |
| ], |
| "year": 2011, |
| "venue": "Journal of machine Learning research", |
| "volume": "12", |
| "issue": "", |
| "pages": "2825--2830", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Fabian Pedregosa, Ga\u00ebl Varoquaux, Alexandre Gram- fort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, et al. 2011. Scikit-learn: Machine learning in python. the Journal of machine Learning research, 12:2825-2830.", |
| "links": null |
| }, |
| "BIBREF34": { |
| "ref_id": "b34", |
| "title": "Linguistic inquiry and word count (LIWC)", |
| "authors": [ |
| { |
| "first": "James", |
| "middle": [], |
| "last": "Pennebaker", |
| "suffix": "" |
| }, |
| { |
| "first": "Martha", |
| "middle": [], |
| "last": "Francis", |
| "suffix": "" |
| }, |
| { |
| "first": "Roger", |
| "middle": [], |
| "last": "Booth", |
| "suffix": "" |
| } |
| ], |
| "year": 2001, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "James Pennebaker, Martha Francis, and Roger Booth. 2001. Linguistic inquiry and word count (LIWC).", |
| "links": null |
| }, |
| "BIBREF35": { |
| "ref_id": "b35", |
| "title": "L'insight neurocognitif dans la schizophr\u00e9nie. L'Enc\u00e9phale", |
| "authors": [ |
| { |
| "first": "St\u00e9phane", |
| "middle": [], |
| "last": "Potvin", |
| "suffix": "" |
| }, |
| { |
| "first": "Ginette", |
| "middle": [], |
| "last": "Aubin", |
| "suffix": "" |
| }, |
| { |
| "first": "Emmanuel", |
| "middle": [], |
| "last": "Stip", |
| "suffix": "" |
| } |
| ], |
| "year": 2017, |
| "venue": "", |
| "volume": "43", |
| "issue": "", |
| "pages": "15--20", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "St\u00e9phane Potvin, Ginette Aubin, and Emmanuel Stip. 2017. L'insight neurocognitif dans la schizophr\u00e9nie. L'Enc\u00e9phale, 43(1):15-20.", |
| "links": null |
| }, |
| "BIBREF36": { |
| "ref_id": "b36", |
| "title": "A comprehensive review of computational methods for automatic prediction of schizophrenia with insight into indigenous populations", |
| "authors": [ |
| { |
| "first": "Randall", |
| "middle": [], |
| "last": "Ratana", |
| "suffix": "" |
| }, |
| { |
| "first": "Hamid", |
| "middle": [], |
| "last": "Sharifzadeh", |
| "suffix": "" |
| }, |
| { |
| "first": "Jamuna", |
| "middle": [], |
| "last": "Krishnan", |
| "suffix": "" |
| }, |
| { |
| "first": "Paul", |
| "middle": [], |
| "last": "Pang", |
| "suffix": "" |
| } |
| ], |
| "year": 2019, |
| "venue": "Frontiers in psychiatry", |
| "volume": "10", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Randall Ratana, Hamid Sharifzadeh, Jamuna Krishnan, and Paul Pang. 2019. A comprehensive review of computational methods for automatic prediction of schizophrenia with insight into indigenous popula- tions. Frontiers in psychiatry, 10:659.", |
| "links": null |
| }, |
| "BIBREF37": { |
| "ref_id": "b37", |
| "title": "Using SDRT to analyze pathological conversations. Logicality, rationality and pragmatic deviances", |
| "authors": [ |
| { |
| "first": "Manuel", |
| "middle": [], |
| "last": "Rebuschi", |
| "suffix": "" |
| }, |
| { |
| "first": "Maxime", |
| "middle": [], |
| "last": "Amblard", |
| "suffix": "" |
| }, |
| { |
| "first": "Michel", |
| "middle": [], |
| "last": "Musiol", |
| "suffix": "" |
| } |
| ], |
| "year": 2014, |
| "venue": "Interdisciplinary Works in Logic, Epistemology, Psychology and Linguistics: Dialogue, Rationality, and Formalism", |
| "volume": "3", |
| "issue": "", |
| "pages": "343--368", |
| "other_ids": { |
| "DOI": [ |
| "10.1007/978-3-319-03044-9_15" |
| ] |
| }, |
| "num": null, |
| "urls": [], |
| "raw_text": "Manuel Rebuschi, Maxime Amblard, and Michel Mu- siol. 2014. Using SDRT to analyze pathological con- versations. Logicality, rationality and pragmatic de- viances. In Manuel Rebuschi, Martine Batt, Ger- hard Heinzmann, Franck Lihoreau, Michel Musiol, and Alain Trognon, editors, Interdisciplinary Works in Logic, Epistemology, Psychology and Linguistics: Dialogue, Rationality, and Formalism, volume 3 of Logic, Argumentation & Reasoning, pages 343 - 368. Springer.", |
| "links": null |
| }, |
| "BIBREF38": { |
| "ref_id": "b38", |
| "title": "Crazy talk: A study of the discourse of schizophrenic speakers", |
| "authors": [ |
| { |
| "first": "Sherry", |
| "middle": [], |
| "last": "Rochester", |
| "suffix": "" |
| } |
| ], |
| "year": 2013, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Sherry Rochester. 2013. Crazy talk: A study of the dis- course of schizophrenic speakers. Springer Science & Business Media.", |
| "links": null |
| }, |
| "BIBREF39": { |
| "ref_id": "b39", |
| "title": "Lexconn: a french lexicon of discourse connectives. Discours. Revue de linguistique, psycholinguistique et informatique. A journal of linguistics, psycholinguistics and computational linguistics", |
| "authors": [ |
| { |
| "first": "Charlotte", |
| "middle": [], |
| "last": "Roze", |
| "suffix": "" |
| }, |
| { |
| "first": "Laurence", |
| "middle": [], |
| "last": "Danlos", |
| "suffix": "" |
| }, |
| { |
| "first": "Philippe", |
| "middle": [], |
| "last": "Muller", |
| "suffix": "" |
| } |
| ], |
| "year": 2012, |
| "venue": "", |
| "volume": "", |
| "issue": "", |
| "pages": "", |
| "other_ids": {}, |
| "num": null, |
| "urls": [], |
| "raw_text": "Charlotte Roze, Laurence Danlos, and Philippe Muller. 2012. Lexconn: a french lexicon of discourse connectives. Discours. Revue de linguistique, psy- cholinguistique et informatique. A journal of linguis- tics, psycholinguistics and computational linguistics, (10).", |
| "links": null |
| } |
| }, |
| "ref_entries": { |
| "FIGREF0": { |
| "text": "based on the de-Max . . . eat . . . apple NOUN . . . VERB . . . NOUN nsubj dobj An example of syntactic relation represented as treelet.", |
| "uris": null, |
| "num": null, |
| "type_str": "figure" |
| }, |
| "FIGREF1": { |
| "text": "Accuracy for all features and window sizes. OCR: Open Class Repair, BC: Backchannel response, Conn.: connectives. W-n: window size.", |
| "uris": null, |
| "num": null, |
| "type_str": "figure" |
| }, |
| "TABREF1": { |
| "text": "Number of documents, speech turns and words per document when varying the window.", |
| "num": null, |
| "html": null, |
| "content": "<table/>", |
| "type_str": "table" |
| }, |
| "TABREF3": { |
| "text": "", |
| "num": null, |
| "html": null, |
| "content": "<table/>", |
| "type_str": "table" |
| }, |
| "TABREF5": { |
| "text": ", c'est \u00e7a[that's right / yeah, right]). At the same time, patients with schizophrenia are correlated with more ambiguous expressions which are also used in non-phatic contexts(je comprends [I understand], bien s\u00fbr [of course], exactement [exactly]), i.e., less BC responses: this supports that the patients are less prone to maintain the conversation. Connectives also give promising results, at best 73.6%. Trend shows that controls use longer connectives (jusqu'\u00e0 ce que [until that], au point de [to the point that]) vs. patients (donc [so], puis [then]). Connectives linked to the present moment are also highly correlated to schizophrenic group (maintenant (que) [now (that)], depuis que [ever since]); this might refer to changes after treatment.", |
| "num": null, |
| "html": null, |
| "content": "<table/>", |
| "type_str": "table" |
| }, |
| "TABREF6": { |
| "text": "Masahito Sakishita, Chihiro Ogawa, Kenji J Tsuchiya, Toshiki Iwabuchi, Taishiro Kishimoto, and Yoshinobu Kano. 2019. Autism spectrum disorder's severity prediction model using utterance features for automatic diagnosis support. In International Workshop on Health Intelligence, pages 83-95. Springer.", |
| "num": null, |
| "html": null, |
| "content": "<table><tr><td colspan=\"2\">K Salzinger. 1979. Ecolinguistics: A radical behavior</td></tr><tr><td>theory approach to language behavior.</td><td/></tr><tr><td colspan=\"2\">Kurt Salzinger and Muriel Hammer. 1963. Some for-</td></tr><tr><td colspan=\"2\">mal characteristics of schizophrenic speech as a mea-</td></tr><tr><td colspan=\"2\">sure of social deviance. Annals of the New York</td></tr><tr><td>Academy of Sciences.</td><td/></tr><tr><td colspan=\"2\">Kurt Salzinger, Stephanie Portnoy, and Richard S Feld-</td></tr><tr><td colspan=\"2\">man. 1964. Verbal behavior of schizophrenic and</td></tr><tr><td colspan=\"2\">normal subjects. Annals of the New York Academy</td></tr><tr><td>of sciences, 105(15):845-860.</td><td/></tr><tr><td colspan=\"2\">Kurt Salzinger, Stephanie Portnoy, David B Pisoni,</td></tr><tr><td>and Richard S Feldman. 1970.</td><td>The immedi-</td></tr><tr><td colspan=\"2\">acy hypothesis and response-produced stimuli in</td></tr><tr><td colspan=\"2\">schizophrenic speech. Journal of Abnormal Psy-</td></tr><tr><td>chology, 76(2):258.</td><td/></tr><tr><td colspan=\"2\">Ivan Sekuli\u0107 and Michael Strube. 2019. Adapting deep</td></tr><tr><td colspan=\"2\">learning methods for mental health prediction on so-</td></tr><tr><td colspan=\"2\">cial media. In Proceedings of the 5th Workshop</td></tr><tr><td colspan=\"2\">on Noisy User-generated Text (W-NUT 2019), pages</td></tr><tr><td>322-327.</td><td/></tr><tr><td colspan=\"2\">Shihao Xu, Zixu Yang, Debsubhra Chakraborty, Yasir</td></tr><tr><td colspan=\"2\">Tahir, Tomasz Maszczyk, Yi Han Victoria Chua,</td></tr><tr><td colspan=\"2\">Justin Dauwels, Daniel Thalmann, Nadia Magnenat</td></tr><tr><td colspan=\"2\">Thalmann, Bhing-Leet Tan, et al. 2019. Automated</td></tr><tr><td colspan=\"2\">lexical analysis of interviews with individuals with</td></tr><tr><td colspan=\"2\">schizophrenia. In Proceedings of the 9th Interna-</td></tr><tr><td colspan=\"2\">tional Workshop on Spoken Dialogue System Tech-</td></tr><tr><td>nology, pages 185-197. Springer.</td><td/></tr></table>", |
| "type_str": "table" |
| }, |
| "TABREF8": { |
| "text": "Open class repair initiators list (French).", |
| "num": null, |
| "html": null, |
| "content": "<table><tr><td>pardon you said</td><td>pardon</td><td>ah pardon you were saying</td></tr><tr><td>excuse-me</td><td>excuse me</td><td>i am sorry</td></tr><tr><td>sorry</td><td>(ah) yes?</td><td>ah really?</td></tr><tr><td>is it true?</td><td>it's euh?</td><td>huh?</td></tr><tr><td>of what</td><td>what is it?</td><td>which means</td></tr><tr><td>euh?</td><td>tell me more</td><td>but still</td></tr></table>", |
| "type_str": "table" |
| }, |
| "TABREF9": { |
| "text": "", |
| "num": null, |
| "html": null, |
| "content": "<table><tr><td colspan=\"3\">: Open class repair initiators list (English trans-</td></tr><tr><td>lation).</td><td/><td/></tr><tr><td>oui</td><td>ouais</td><td>ouais voil\u00e0</td></tr><tr><td>oui c'est \u00e7a</td><td>oui bah oui</td><td>oui... forc\u00e9ment</td></tr><tr><td>bah ouais</td><td>hum (hum)</td><td>muh mmh</td></tr><tr><td>mmh/mmhh</td><td>d'accord</td><td>ok</td></tr><tr><td>voil\u00e0</td><td>c'est \u00e7a</td><td>c'est vrai</td></tr><tr><td>c'est s\u00fbr</td><td>\u00e7a c'est clair</td><td>eh bien s\u00fbr</td></tr><tr><td>carr\u00e9ment</td><td>bien s\u00fbr</td><td>super</td></tr><tr><td>ok... bon</td><td>d'accord \u00e7a marche</td><td>certes</td></tr><tr><td>mais hein</td><td>je comprends</td><td>vraiment</td></tr><tr><td>bien</td><td>bon</td><td>tr\u00e8s bien</td></tr><tr><td>quand m\u00eame</td><td>tout \u00e0 fait</td><td>certainement</td></tr><tr><td>exactement</td><td>tant mieux</td><td>oh</td></tr><tr><td>ah</td><td>ben</td><td>alors ben</td></tr><tr><td>ah d'accord</td><td>ah \u00e7a euh</td><td>eh bah c'est bien</td></tr></table>", |
| "type_str": "table" |
| }, |
| "TABREF10": { |
| "text": "backchannel response list (French).", |
| "num": null, |
| "html": null, |
| "content": "<table><tr><td>yes</td><td>yeah</td><td>yeah that's it</td></tr><tr><td>yes that's it</td><td>yes euh yes</td><td>yes... for sure</td></tr><tr><td>euh yeah</td><td>hum (hum)</td><td>muh mmh</td></tr><tr><td>mmh/mmhh</td><td>okay</td><td>ok</td></tr><tr><td>that's it</td><td>that's it</td><td>that's true</td></tr><tr><td>(yes) (for) sure</td><td>that's clear/clearly/definitely</td><td>eh of course</td></tr><tr><td>completely</td><td>of course</td><td>super</td></tr><tr><td>ok... then</td><td>all right</td><td>indeed/yes</td></tr><tr><td>but hein</td><td>i understand</td><td>really</td></tr><tr><td>good</td><td>well</td><td>very good</td></tr><tr><td>still</td><td>exactly</td><td>certainly/sure</td></tr><tr><td>exactly</td><td>all the better/so much the better</td><td>oh</td></tr><tr><td>ah</td><td>well</td><td>so... well</td></tr><tr><td>ah okay</td><td>ah (this) euh</td><td>eh well that's good</td></tr></table>", |
| "type_str": "table" |
| }, |
| "TABREF11": { |
| "text": "", |
| "num": null, |
| "html": null, |
| "content": "<table><tr><td>: backchannel response list (English transla-</td></tr><tr><td>tion).</td></tr></table>", |
| "type_str": "table" |
| } |
| } |
| } |
| } |
