TCTao Chen

Tao Chen

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Tao Chen has a total of 57 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Pathways of Thoughts: Multi-Directional Thinking for Long-form Personalized Question Answering

    by Alireza Salemi, Cheng Li, Mingyang Zhang, Qiaozhu Mei, Zhuowan Li, Spurthi Amba Hombaiah, Weize Kong, Tao Chen, Hamed Zamani, Michael Bendersky

  • Open Access

    End-to-End Query Term Weighting

    by Karan Samel, Cheng Li, Weize Kong, Tao Chen, Mingyang Zhang, Shaleen Gupta, Swaraj Khadanga, Wensong Xu, Xingyu Wang, Kashyap Kolipaka, et al.

    Bag-of-words based lexical retrieval systems are still the most commonly used methods for real-world search applications. Recently deep learning methods have shown promising results to improve this retrieval performance but are expensive to run in an online fashion, non-trivial to integrate into existing production systems, and might not generalize well in out-of-domain retrieval scenarios. Instead, we build on top of lexical retrievers by proposing a Term Weighting BERT (TW-BERT) model. TW-BERT learns to predict the weight for individual n-gram (e.g., uni-grams and bi-grams) query input terms. These inferred weights and terms can be used directly by a retrieval system to perform a query search. To optimize these term weights, TW-BERT incorporates the scoring function used by the search engine, such as BM25, to score query-document pairs. Given sample query-document pairs we can compute a ranking loss over these matching scores, optimizing the learned query term weights in an end-to-end fashion. Aligning TW-BERT with search engine scorers minimizes the changes needed to integrate it into existing production applications, whereas existing deep learning based search methods would require further infrastructure optimization and hardware requirements. The learned weights can be easily utilized by standard lexical retrievers and by other retrieval techniques such as query expansion. We show that TW-BERT improves retrieval performance over strong term weighting baselines within MSMARCO and in out-of-domain retrieval on TREC datasets.


  • Real-time Personalized Ranking and Recommendation System for Linear TV: A Dual Dynamic Queue Approach

    by Ning Xu, Tao Chen

    The increasing number of channels on linear TV has amplified the challenge of efficient content discovery for viewers. This paper presents an innovative, real-time personalized ranking and recommendation system designed to address this problem. The objective was to enhance viewer satisfaction and interaction through a personalized, dynamic channel surfing experience. Our approach uses a dual dynamic queue system - the Dynamic History Channel Queue (DHCQ) and the Dynamic Future Channel Queue (DFCQ), each serving a unique purpose in managing the viewer's channel interaction. — Advanced deep learning models reinforce these queues’ dynamism by generating ‘global’ and ‘local’ content embeddings and ‘user’ embeddings. These embeddings are used to provide real-time updates, considering both timestamps and the time-sensitivities of the content and viewer. A ‘look- ahead’ feature was integrated to account for future content on each channel, adding another layer of personalization. — Preliminary user feedback highlighted a strong interest in this new form of personalized channel surfing, with a majority of respondents indicating a preference for our system over traditional channel navigation methods. The results show that our proposed solution could potentially change the way users interact with the ever-growing number of channels on linear TV.

  • A Management Technology of Electrical User Data Label Based on Artificial Intelligence Identification

    by Xiao Cheng, Jinyang Jiang, Fei Xiang, Xin Yin, Dinghao Liu, Tao Chen

    With the rapid development of artificial intelligence technology, it has been applied in various businesses of enterprises. State Grid Corporation of China generates massive amounts of electricity user data in its regular business. The processing and application of such data mainly rely on manual extraction, classification, and visual processing. This method has low efficiency, high labor cost, and low accuracy. Therefore, the technology of artificial intelligence recognition technology has become an ideal new alternative. This paper is based on the Globe (Global Vectors for Word Representation, Globe) model to identify and represent electrical customer data and label different types of data, and finally build a new set of electrical user data label management technology. It realizes the efficient storage and management of massive electrical user data.


  • Identifying vulnerable older adult populations by contextualizing geriatric syndrome information in clinical notes of electronic health records

    by Tao Chen, Mark Dredze, Jonathan P Weiner, Hadi Kharrazi

    Geriatric syndromes such as functional disability and lack of social support are often not encoded in electronic health records (EHRs), thus obscuring the identification of vulnerable older adults in need of additional medical and social services. In this study, we automatically identify vulnerable older adult patients with geriatric syndrome based on clinical notes extracted from an EHR system, and demonstrate how contextual information can improve the process. We propose a novel end-to-end neural architecture to identify sentences that contain geriatric syndromes. Our model learns a representation of the sentence and augments it with contextual information: surrounding sentences, the entire clinical document, and the diagnosis codes associated with the document. We trained our system on annotated notes from 85 patients, tuned the model on another 50 patients, and evaluated its performance on the rest, 50 patients. Contextual information improved classification, with the most effective context coming from the surrounding sentences. At sentence level, our best performing model achieved a micro-F1 of 0.605, significantly outperforming context-free baselines. At patient level, our best model achieved a micro-F1 of 0.843. Our solution can be used to expand the identification of vulnerable older adults with geriatric syndromes. Since functional and social factors are often not captured by diagnosis codes in EHRs, the automatic identification of the geriatric syndrome can reduce disparities by ensuring consistent care across the older adult population. EHR free-text can be used to identify vulnerable older adults with a range of geriatric syndromes.


  • Open Access

    Extraction of Geriatric Syndromes From Electronic Health Record Clinical Notes: Assessment of Statistical Natural Language Processing Methods

    by Tao Chen, Mark Dredze, Jonathan P Weiner, Leilani Hernandez, Joe Kimura, Hadi Kharrazi

    Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). Background: Geriatric syndromes in older adults are associated with adverse outcomes. However, despite being reported in clinical notes these syndromes are often poorly captured by diagnostic codes in the structured fields of electronic health records (EHRs) or administrative records. Objective: We aim to automatically determine if a patient has any geriatric syndromes by mining the free text of associated EHR clinical notes. We assessed which statistical natural language processing (NLP) techniques are most effective. Methods: We applied Conditional Random Fields (CRFs), a widely used machine learning algorithm, to identify each of 10 geriatric syndrome constructs in a clinical note. We assessed three sets of features/attributes for CRF operations: a base set, enhanced token, and contextual features. We trained the CRF on 3901 manually annotated notes from 85 patients, tuned the CRF on a validation set of 50 patients, and evaluated it on 50 held-out test patients. These notes were from a group of US Medicare (over 65) patients enrolled in a "Medicare-Advantage" HMO and cared for by a large group practice in Massachusetts. Results: A final feature set was formed through comprehensive feature ablation experiments. The final CRF model performed well at patient-level determination (macro-F1=0.834, micro-F1=0.851); however, performance varied by construct. For example, at phrase-partial evaluation, CRF worked well on constructs like absence of fecal control (F1=0.857) and vision impairment (F1=0.798), but poorly on malnutrition (F1=0.155), weight loss (F1=0.394) and severe urinary control issues (F1=0.532). Errors were primarily due to previously unobserved words (out-of-vocabulary) and a lack of context. Conclusions: This study shows that statistical NLP can be used to identify geriatric syndromes from EHR-extracted clinical notes. This creates new opportunities to identify patients with geriatric syndromes and study their health outcomes.


  • Open Access

    Extraction of Geriatric Syndromes From Electronic Health Record Clinical Notes: Assessment of Statistical Natural Language Processing Methods (Preprint)

    by Tao Chen, Mark Dredze, Jonathan P Weiner, Leilani Hernandez, Joe Kimura, Hadi Kharrazi

    BACKGROUND Geriatric syndromes in older adults are associated with adverse outcomes. However, despite being reported in clinical notes, these syndromes are often poorly captured by diagnostic codes in the structured fields of electronic health records (EHRs) or administrative records. OBJECTIVE We aim to automatically determine if a patient has any geriatric syndromes by mining the free text of associated EHR clinical notes. We assessed which statistical natural language processing (NLP) techniques are most effective. METHODS We applied conditional random fields (CRFs), a widely used machine learning algorithm, to identify each of 10 geriatric syndrome constructs in a clinical note. We assessed three sets of features and attributes for CRF operations: a base set, enhanced token, and contextual features. We trained the CRF on 3901 manually annotated notes from 85 patients, tuned the CRF on a validation set of 50 patients, and evaluated it on 50 held-out test patients. These notes were from a group of US Medicare patients over 65 years of age enrolled in a Medicare Advantage Health Maintenance Organization and cared for by a large group practice in Massachusetts. RESULTS A final feature set was formed through comprehensive feature ablation experiments. The final CRF model performed well at patient-level determination (macroaverage F1=0.834, microaverage F1=0.851); however, performance varied by construct. For example, at phrase-partial evaluation, the CRF model worked well on constructs such as absence of fecal control (F1=0.857) and vision impairment (F1=0.798) but poorly on malnutrition (F1=0.155), weight loss (F1=0.394), and severe urinary control issues (F1=0.532). Errors were primarily due to previously unobserved words (ie, out-of-vocabulary) and a lack of context. CONCLUSIONS This study shows that statistical NLP can be used to identify geriatric syndromes from EHR-extracted clinical notes. This creates new opportunities to identify patients with geriatric syndromes and study their health outcomes.

  • Top Cited Papers

    Weaponized Health Communication: Twitter Bots and Russian Trolls Amplify the Vaccine Debate

    by David A. Broniatowski, Amelia M. Jamison, SiHua Qi, Lulwah AlKulaib, Tao Chen, Adrian Benton, Sandra C. Quinn, Mark Dredze

    Objectives. To understand how Twitter bots and trolls (“bots”) promote online health content. Methods. We compared bots’ to average users’ rates of vaccine-relevant messages, which we collected online from July 2014 through September 2017. We estimated the likelihood that users were bots, comparing proportions of polarized and antivaccine tweets across user types. We conducted a content analysis of a Twitter hashtag associated with Russian troll activity. Results. Compared with average users, Russian trolls (χ2(1) = 102.0; P < .001), sophisticated bots (χ2(1) = 28.6; P < .001), and “content polluters” (χ2(1) = 7.0; P < .001) tweeted about vaccination at higher rates. Whereas content polluters posted more antivaccine content (χ2(1) = 11.18; P < .001), Russian trolls amplified both sides. Unidentifiable accounts were more polarized (χ2(1) = 12.1; P < .001) and antivaccine (χ2(1) = 35.9; P < .001). Analysis of the Russian troll hashtag showed that its messages were more political and divisive. Conclusions. Whereas bots that spread malware and unsolicited content disseminated antivaccine messages, Russian trolls promoted discord. Accounts masquerading as legitimate users create false equivalency, eroding public consensus on vaccination. Public Health Implications. Directly confronting vaccine skeptics enables bots to legitimize the vaccine debate. More research is needed to determine how best to combat bot-driven content.


  • Open Access

    Discordance Between Human Papillomavirus Twitter Images and Disparities in Human Papillomavirus Risk and Disease in the United States: Mixed-Methods Analysis

    by Yuki Lama, Tao Chen, Mark Dredze, Amelia Jamison, Sandra Crouse Quinn, David A Broniatowski

    Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). Background: Racial and ethnic minorities are disproportionately affected by human papillomavirus (HPV) related cancers. HPV vaccination remains low among these populations. Given the importance of social media platforms for health communication, we examined HPV images on Twitter. We explored inconsistencies between the demographics represented in HPV images and the populations who experience the greatest burden of HPV-related disease. Objective: Our study aims to (1) observe whether HPV images on Twitter reflect actual burden of disease by select demographics and (2) determine to what extent Twitter accounts utilized images that reflect burden of disease in their health communication messages. Methods: We identified 456 image tweets about HPV that contained faces which were posted by US users between Nov 11, 2014 and Aug 8, 2016. We identified images containing at least one human face and utilized Face++ software to automatically extract the gender, age, and race of each face. We next categorized the source accounts of these tweets into three types: government (12.8%), organizations (54.0%), and individual (33.2%) to examine how images varied by message source. Results: Findings reflect racial demographics of the US population, but not disease burden (65.2% White faces, 11.5% Black faces, 5.8% Asian faces, and 17.5% racially ambiguous faces). Gender disparities are evident in the image faces: 71.7% represent female faces, whereas only 27.9% represent male faces. Among the 11-26 age group recommended to receive HPV vaccine, HPV images contain more female only faces (34.3%) than males (6.0%); remainder of images include both male and female faces (59.3%). Gender and racial disparities were present across different image sources. Faces from government sources were more likely to depict females (n=44) in comparison to males (n=16). Of the male faces, 80% of youth and 100% of adults were White. News organization sources depicted high proportion of White faces (96.6% of female youth, 100% of adult males). Conclusions: Our results show critical differences between the demographics reflected in HPV images and the actual burden of disease. Racial minorities are less likely to appear in HPV images despite higher rates of HPV incidence. Health communication efforts need to better represent populations at risk if we seek to reduce disparities in HPV infection


  • Open Access

    Vaccine Images on Twitter: Analysis of What Images are Shared

    by Tao Chen, Mark Dredze

    Visual imagery plays a key role in health communication; however, there is little understanding of what aspects of vaccine-related images make them effective communication aids. Twitter, a popular venue for discussions related to vaccination, provides numerous images that are shared with tweets. The objectives of this study were to understand how images are used in vaccine-related tweets and provide guidance with respect to the characteristics of vaccine-related images that correlate with the higher likelihood of being retweeted. We collected more than one million vaccine image messages from Twitter and characterized various properties of these images using automated image analytics. We fit a logistic regression model to predict whether or not a vaccine image tweet was retweeted, thus identifying characteristics that correlate with a higher likelihood of being shared. For comparison, we built similar models for the sharing of vaccine news on Facebook and for general image tweets. Most vaccine-related images are duplicates (125,916/237,478; 53.02%) or taken from other sources, not necessarily created by the author of the tweet. Almost half of the images contain embedded text, and many include images of people and syringes. The visual content is highly correlated with a tweet's textual topics. Vaccine image tweets are twice as likely to be shared as nonimage tweets. The sentiment of an image and the objects shown in the image were the predictive factors in determining whether an image was retweeted. We are the first to study vaccine images on Twitter. Our findings suggest future directions for the study and use of vaccine imagery and may inform communication strategies around vaccination. Furthermore, our study demonstrates an effective study methodology for image analysis.