Pub Date : 2023-01-23DOI: 10.48550/arXiv.2301.09728
Arian Askari, Amin Abolghasemi, G. Pasi, Wessel Kraaij, S. Verberne
In this paper we propose a novel approach for combining first-stage lexical retrieval models and Transformer-based re-rankers: we inject the relevance score of the lexical model as a token in the middle of the input of the cross-encoder re-ranker. It was shown in prior work that interpolation between the relevance score of lexical and BERT-based re-rankers may not consistently result in higher effectiveness. Our idea is motivated by the finding that BERT models can capture numeric information. We compare several representations of the BM25 score and inject them as text in the input of four different cross-encoders. We additionally analyze the effect for different query types, and investigate the effectiveness of our method for capturing exact matching relevance. Evaluation on the MSMARCO Passage collection and the TREC DL collections shows that the proposed method significantly improves over all cross-encoder re-rankers as well as the common interpolation methods. We show that the improvement is consistent for all query types. We also find an improvement in exact matching capabilities over both BM25 and the cross-encoders. Our findings indicate that cross-encoder re-rankers can efficiently be improved without additional computational burden and extra steps in the pipeline by explicitly adding the output of the first-stage ranker to the model input, and this effect is robust for different models and query types.
{"title":"Injecting the BM25 Score as Text Improves BERT-Based Re-rankers","authors":"Arian Askari, Amin Abolghasemi, G. Pasi, Wessel Kraaij, S. Verberne","doi":"10.48550/arXiv.2301.09728","DOIUrl":"https://doi.org/10.48550/arXiv.2301.09728","url":null,"abstract":"In this paper we propose a novel approach for combining first-stage lexical retrieval models and Transformer-based re-rankers: we inject the relevance score of the lexical model as a token in the middle of the input of the cross-encoder re-ranker. It was shown in prior work that interpolation between the relevance score of lexical and BERT-based re-rankers may not consistently result in higher effectiveness. Our idea is motivated by the finding that BERT models can capture numeric information. We compare several representations of the BM25 score and inject them as text in the input of four different cross-encoders. We additionally analyze the effect for different query types, and investigate the effectiveness of our method for capturing exact matching relevance. Evaluation on the MSMARCO Passage collection and the TREC DL collections shows that the proposed method significantly improves over all cross-encoder re-rankers as well as the common interpolation methods. We show that the improvement is consistent for all query types. We also find an improvement in exact matching capabilities over both BM25 and the cross-encoders. Our findings indicate that cross-encoder re-rankers can efficiently be improved without additional computational burden and extra steps in the pipeline by explicitly adding the output of the first-stage ranker to the model input, and this effect is robust for different models and query types.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126906690","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-19DOI: 10.48550/arXiv.2301.08062
Mehmet Deniz Turkmen, Matthew Lease, Mucahid Kutlu
In evaluation campaigns, participants often explore variations of popular, state-of-the-art baselines as a low-risk strategy to achieve competitive results. While effective, this can lead to local"hill climbing"rather than more radical and innovative departure from standard methods. Moreover, if many participants build on similar baselines, the overall diversity of approaches considered may be limited. In this work, we propose a new class of IR evaluation metrics intended to promote greater diversity of approaches in evaluation campaigns. Whereas traditional IR metrics focus on user experience, our two"innovation"metrics instead reward exploration of more divergent, higher-risk strategies finding relevant documents missed by other systems. Experiments on four TREC collections show that our metrics do change system rankings by rewarding systems that find such rare, relevant documents. This result is further supported by a controlled, synthetic data experiment, and a qualitative analysis. In addition, we show that our metrics achieve higher evaluation stability and discriminative power than the standard metrics we modify. To support reproducibility, we share our source code.
{"title":"New Metrics to Encourage Innovation and Diversity in Information Retrieval Approaches","authors":"Mehmet Deniz Turkmen, Matthew Lease, Mucahid Kutlu","doi":"10.48550/arXiv.2301.08062","DOIUrl":"https://doi.org/10.48550/arXiv.2301.08062","url":null,"abstract":"In evaluation campaigns, participants often explore variations of popular, state-of-the-art baselines as a low-risk strategy to achieve competitive results. While effective, this can lead to local\"hill climbing\"rather than more radical and innovative departure from standard methods. Moreover, if many participants build on similar baselines, the overall diversity of approaches considered may be limited. In this work, we propose a new class of IR evaluation metrics intended to promote greater diversity of approaches in evaluation campaigns. Whereas traditional IR metrics focus on user experience, our two\"innovation\"metrics instead reward exploration of more divergent, higher-risk strategies finding relevant documents missed by other systems. Experiments on four TREC collections show that our metrics do change system rankings by rewarding systems that find such rare, relevant documents. This result is further supported by a controlled, synthetic data experiment, and a qualitative analysis. In addition, we show that our metrics achieve higher evaluation stability and discriminative power than the standard metrics we modify. To support reproducibility, we share our source code.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127068078","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-19DOI: 10.48550/arXiv.2301.08006
Jorge Gab'in, M. E. Ares, Javier Parapar
Nowadays, search engine users commonly rely on query suggestions to improve their initial inputs. Current systems are very good at recommending lexical adaptations or spelling corrections to users' queries. However, they often struggle to suggest semantically related keywords given a user's query. The construction of a detailed query is crucial in some tasks, such as legal retrieval or academic search. In these scenarios, keyword suggestion methods are critical to guide the user during the query formulation. This paper proposes two novel models for the keyword suggestion task trained on scientific literature. Our techniques adapt the architecture of Word2Vec and FastText to generate keyword embeddings by leveraging documents' keyword co-occurrence. Along with these models, we also present a specially tailored negative sampling approach that exploits how keywords appear in academic publications. We devise a ranking-based evaluation methodology following both known-item and ad-hoc search scenarios. Finally, we evaluate our proposals against the state-of-the-art word and sentence embedding models showing considerable improvements over the baselines for the tasks.
{"title":"Keyword Embeddings for Query Suggestion","authors":"Jorge Gab'in, M. E. Ares, Javier Parapar","doi":"10.48550/arXiv.2301.08006","DOIUrl":"https://doi.org/10.48550/arXiv.2301.08006","url":null,"abstract":"Nowadays, search engine users commonly rely on query suggestions to improve their initial inputs. Current systems are very good at recommending lexical adaptations or spelling corrections to users' queries. However, they often struggle to suggest semantically related keywords given a user's query. The construction of a detailed query is crucial in some tasks, such as legal retrieval or academic search. In these scenarios, keyword suggestion methods are critical to guide the user during the query formulation. This paper proposes two novel models for the keyword suggestion task trained on scientific literature. Our techniques adapt the architecture of Word2Vec and FastText to generate keyword embeddings by leveraging documents' keyword co-occurrence. Along with these models, we also present a specially tailored negative sampling approach that exploits how keywords appear in academic publications. We devise a ranking-based evaluation methodology following both known-item and ad-hoc search scenarios. Finally, we evaluate our proposals against the state-of-the-art word and sentence embedding models showing considerable improvements over the baselines for the tasks.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124340407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-14DOI: 10.48550/arXiv.2301.05863
Fatima Haouari, Tamer Elsayed
A myriad of studies addressed the problem of rumor verification in Twitter by either utilizing evidence from the propagation networks or external evidence from the Web. However, none of these studies exploited evidence from trusted authorities. In this paper, we define the task of detecting the stance of authorities towards rumors in tweets, i.e., whether a tweet from an authority agrees, disagrees, or is unrelated to the rumor. We believe the task is useful to augment the sources of evidence utilized by existing rumor verification systems. We construct and release the first Authority STance towards Rumors (AuSTR) dataset, where evidence is retrieved from authority timelines in Arabic Twitter. Due to the relatively limited size of our dataset, we study the usefulness of existing datasets for stance detection in our task. We show that existing datasets are somewhat useful for the task; however, they are clearly insufficient, which motivates the need to augment them with annotated data constituting stance of authorities from Twitter.
{"title":"Detecting Stance of Authorities towards Rumors in Arabic Tweets: A Preliminary Study","authors":"Fatima Haouari, Tamer Elsayed","doi":"10.48550/arXiv.2301.05863","DOIUrl":"https://doi.org/10.48550/arXiv.2301.05863","url":null,"abstract":"A myriad of studies addressed the problem of rumor verification in Twitter by either utilizing evidence from the propagation networks or external evidence from the Web. However, none of these studies exploited evidence from trusted authorities. In this paper, we define the task of detecting the stance of authorities towards rumors in tweets, i.e., whether a tweet from an authority agrees, disagrees, or is unrelated to the rumor. We believe the task is useful to augment the sources of evidence utilized by existing rumor verification systems. We construct and release the first Authority STance towards Rumors (AuSTR) dataset, where evidence is retrieved from authority timelines in Arabic Twitter. Due to the relatively limited size of our dataset, we study the usefulness of existing datasets for stance detection in our task. We show that existing datasets are somewhat useful for the task; however, they are clearly insufficient, which motivates the need to augment them with annotated data constituting stance of authorities from Twitter.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125644301","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-14DOI: 10.48550/arXiv.2301.05944
Giacomo Balloccu, Ludovico Boratto, Christian Cancedda, G. Fenu, M. Marras
Path reasoning is a notable recommendation approach that models high-order user-product relations, based on a Knowledge Graph (KG). This approach can extract reasoning paths between recommended products and already experienced products and, then, turn such paths into textual explanations for the user. Unfortunately, evaluation protocols in this field appear heterogeneous and limited, making it hard to contextualize the impact of the existing methods. In this paper, we replicated three state-of-the-art relevant path reasoning recommendation methods proposed in top-tier conferences. Under a common evaluation protocol, based on two public data sets and in comparison with other knowledge-aware methods, we then studied the extent to which they meet recommendation utility and beyond objectives, explanation quality, and consumer and provider fairness. Our study provides a picture of the progress in this field, highlighting open issues and future directions. Source code: url{https://github.com/giacoballoccu/rep-path-reasoning-recsys}.
{"title":"Knowledge is Power, Understanding is Impact: Utility and Beyond Goals, Explanation Quality, and Fairness in Path Reasoning Recommendation","authors":"Giacomo Balloccu, Ludovico Boratto, Christian Cancedda, G. Fenu, M. Marras","doi":"10.48550/arXiv.2301.05944","DOIUrl":"https://doi.org/10.48550/arXiv.2301.05944","url":null,"abstract":"Path reasoning is a notable recommendation approach that models high-order user-product relations, based on a Knowledge Graph (KG). This approach can extract reasoning paths between recommended products and already experienced products and, then, turn such paths into textual explanations for the user. Unfortunately, evaluation protocols in this field appear heterogeneous and limited, making it hard to contextualize the impact of the existing methods. In this paper, we replicated three state-of-the-art relevant path reasoning recommendation methods proposed in top-tier conferences. Under a common evaluation protocol, based on two public data sets and in comparison with other knowledge-aware methods, we then studied the extent to which they meet recommendation utility and beyond objectives, explanation quality, and consumer and provider fairness. Our study provides a picture of the progress in this field, highlighting open issues and future directions. Source code: url{https://github.com/giacoballoccu/rep-path-reasoning-recsys}.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117182679","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-13DOI: 10.48550/arXiv.2301.05453
Ana-Maria Bucur, Adrian Cosma, Paolo Rosso, Liviu P. Dinu
Depression detection from user-generated content on the internet has been a long-lasting topic of interest in the research community, providing valuable screening tools for psychologists. The ubiquitous use of social media platforms lays out the perfect avenue for exploring mental health manifestations in posts and interactions with other users. Current methods for depression detection from social media mainly focus on text processing, and only a few also utilize images posted by users. In this work, we propose a flexible time-enriched multimodal transformer architecture for detecting depression from social media posts, using pretrained models for extracting image and text embeddings. Our model operates directly at the user-level, and we enrich it with the relative time between posts by using time2vec positional embeddings. Moreover, we propose another model variant, which can operate on randomly sampled and unordered sets of posts to be more robust to dataset noise. We show that our method, using EmoBERTa and CLIP embeddings, surpasses other methods on two multimodal datasets, obtaining state-of-the-art results of 0.931 F1 score on a popular multimodal Twitter dataset, and 0.902 F1 score on the only multimodal Reddit dataset.
{"title":"It's Just a Matter of Time: Detecting Depression with Time-Enriched Multimodal Transformers","authors":"Ana-Maria Bucur, Adrian Cosma, Paolo Rosso, Liviu P. Dinu","doi":"10.48550/arXiv.2301.05453","DOIUrl":"https://doi.org/10.48550/arXiv.2301.05453","url":null,"abstract":"Depression detection from user-generated content on the internet has been a long-lasting topic of interest in the research community, providing valuable screening tools for psychologists. The ubiquitous use of social media platforms lays out the perfect avenue for exploring mental health manifestations in posts and interactions with other users. Current methods for depression detection from social media mainly focus on text processing, and only a few also utilize images posted by users. In this work, we propose a flexible time-enriched multimodal transformer architecture for detecting depression from social media posts, using pretrained models for extracting image and text embeddings. Our model operates directly at the user-level, and we enrich it with the relative time between posts by using time2vec positional embeddings. Moreover, we propose another model variant, which can operate on randomly sampled and unordered sets of posts to be more robust to dataset noise. We show that our method, using EmoBERTa and CLIP embeddings, surpasses other methods on two multimodal datasets, obtaining state-of-the-art results of 0.931 F1 score on a popular multimodal Twitter dataset, and 0.902 F1 score on the only multimodal Reddit dataset.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130833961","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-13DOI: 10.48550/arXiv.2301.05494
I. Baris Schlicht, Lucie Flek, Paolo Rosso
Check-worthiness detection is the task of identifying claims, worthy to be investigated by fact-checkers. Resource scarcity for non-world languages and model learning costs remain major challenges for the creation of models supporting multilingual check-worthiness detection. This paper proposes cross-training adapters on a subset of world languages, combined by adapter fusion, to detect claims emerging globally in multiple languages. (1) With a vast number of annotators available for world languages and the storage-efficient adapter models, this approach is more cost efficient. Models can be updated more frequently and thus stay up-to-date. (2) Adapter fusion provides insights and allows for interpretation regarding the influence of each adapter model on a particular language. The proposed solution often outperformed the top multilingual approaches in our benchmark tasks.
{"title":"Multilingual Detection of Check-Worthy Claims using World Languages and Adapter Fusion","authors":"I. Baris Schlicht, Lucie Flek, Paolo Rosso","doi":"10.48550/arXiv.2301.05494","DOIUrl":"https://doi.org/10.48550/arXiv.2301.05494","url":null,"abstract":"Check-worthiness detection is the task of identifying claims, worthy to be investigated by fact-checkers. Resource scarcity for non-world languages and model learning costs remain major challenges for the creation of models supporting multilingual check-worthiness detection. This paper proposes cross-training adapters on a subset of world languages, combined by adapter fusion, to detect claims emerging globally in multiple languages. (1) With a vast number of annotators available for world languages and the storage-efficient adapter models, this approach is more cost efficient. Models can be updated more frequently and thus stay up-to-date. (2) Adapter fusion provides insights and allows for interpretation regarding the influence of each adapter model on a particular language. The proposed solution often outperformed the top multilingual approaches in our benchmark tasks.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128927975","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-13DOI: 10.48550/arXiv.2301.05508
Gustavo Penha, C. Hauff
A number of learned sparse and dense retrieval approaches have recently been proposed and proven effective in tasks such as passage retrieval and document retrieval. In this paper we analyze with a replicability study if the lessons learned generalize to the retrieval of responses for dialogues, an important task for the increasingly popular field of conversational search. Unlike passage and document retrieval where documents are usually longer than queries, in response ranking for dialogues the queries (dialogue contexts) are often longer than the documents (responses). Additionally, dialogues have a particular structure, i.e. multiple utterances by different users. With these differences in mind, we here evaluate how generalizable the following major findings from previous works are: (F1) query expansion outperforms a no-expansion baseline; (F2) document expansion outperforms a no-expansion baseline; (F3) zero-shot dense retrieval underperforms sparse baselines; (F4) dense retrieval outperforms sparse baselines; (F5) hard negative sampling is better than random sampling for training dense models. Our experiments -- based on three different information-seeking dialogue datasets -- reveal that four out of five findings (F2-F5) generalize to our domain
{"title":"Do the Findings of Document and Passage Retrieval Generalize to the Retrieval of Responses for Dialogues?","authors":"Gustavo Penha, C. Hauff","doi":"10.48550/arXiv.2301.05508","DOIUrl":"https://doi.org/10.48550/arXiv.2301.05508","url":null,"abstract":"A number of learned sparse and dense retrieval approaches have recently been proposed and proven effective in tasks such as passage retrieval and document retrieval. In this paper we analyze with a replicability study if the lessons learned generalize to the retrieval of responses for dialogues, an important task for the increasingly popular field of conversational search. Unlike passage and document retrieval where documents are usually longer than queries, in response ranking for dialogues the queries (dialogue contexts) are often longer than the documents (responses). Additionally, dialogues have a particular structure, i.e. multiple utterances by different users. With these differences in mind, we here evaluate how generalizable the following major findings from previous works are: (F1) query expansion outperforms a no-expansion baseline; (F2) document expansion outperforms a no-expansion baseline; (F3) zero-shot dense retrieval underperforms sparse baselines; (F4) dense retrieval outperforms sparse baselines; (F5) hard negative sampling is better than random sampling for training dense models. Our experiments -- based on three different information-seeking dialogue datasets -- reveal that four out of five findings (F2-F5) generalize to our domain","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130169842","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-11DOI: 10.48550/arXiv.2301.04366
Paul Lerner, O. Ferret, C. Guinaudeau
We present a new pre-training method, Multimodal Inverse Cloze Task, for Knowledge-based Visual Question Answering about named Entities (KVQAE). KVQAE is a recently introduced task that consists in answering questions about named entities grounded in a visual context using a Knowledge Base. Therefore, the interaction between the modalities is paramount to retrieve information and must be captured with complex fusion models. As these models require a lot of training data, we design this pre-training task from existing work in textual Question Answering. It consists in considering a sentence as a pseudo-question and its context as a pseudo-relevant passage and is extended by considering images near texts in multimodal documents. Our method is applicable to different neural network architectures and leads to a 9% relative-MRR and 15% relative-F1 gain for retrieval and reading comprehension, respectively, over a no-pre-training baseline.
{"title":"Multimodal Inverse Cloze Task for Knowledge-based Visual Question Answering","authors":"Paul Lerner, O. Ferret, C. Guinaudeau","doi":"10.48550/arXiv.2301.04366","DOIUrl":"https://doi.org/10.48550/arXiv.2301.04366","url":null,"abstract":"We present a new pre-training method, Multimodal Inverse Cloze Task, for Knowledge-based Visual Question Answering about named Entities (KVQAE). KVQAE is a recently introduced task that consists in answering questions about named entities grounded in a visual context using a Knowledge Base. Therefore, the interaction between the modalities is paramount to retrieve information and must be captured with complex fusion models. As these models require a lot of training data, we design this pre-training task from existing work in textual Question Answering. It consists in considering a sentence as a pseudo-question and its context as a pseudo-relevant passage and is extended by considering images near texts in multimodal documents. Our method is applicable to different neural network architectures and leads to a 9% relative-MRR and 15% relative-F1 gain for retrieval and reading comprehension, respectively, over a no-pre-training baseline.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114365908","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-01-11DOI: 10.48550/arXiv.2301.04339
Mozhgan Talebpour, A. G. S. D. Herrera, Shoaib Jameel
Contextualised word vectors obtained via pre-trained language models encode a variety of knowledge that has already been exploited in applications. Complementary to these language models are probabilistic topic models that learn thematic patterns from the text. Recent work has demonstrated that conducting clustering on the word-level contextual representations from a language model emulates word clusters that are discovered in latent topics of words from Latent Dirichlet Allocation. The important question is how such topical word clusters are automatically formed, through clustering, in the language model when it has not been explicitly designed to model latent topics. To address this question, we design different probe experiments. Using BERT and DistilBERT, we find that the attention framework plays a key role in modelling such word topic clusters. We strongly believe that our work paves way for further research into the relationships between probabilistic topic models and pre-trained language models.
{"title":"Topics in Contextualised Attention Embeddings","authors":"Mozhgan Talebpour, A. G. S. D. Herrera, Shoaib Jameel","doi":"10.48550/arXiv.2301.04339","DOIUrl":"https://doi.org/10.48550/arXiv.2301.04339","url":null,"abstract":"Contextualised word vectors obtained via pre-trained language models encode a variety of knowledge that has already been exploited in applications. Complementary to these language models are probabilistic topic models that learn thematic patterns from the text. Recent work has demonstrated that conducting clustering on the word-level contextual representations from a language model emulates word clusters that are discovered in latent topics of words from Latent Dirichlet Allocation. The important question is how such topical word clusters are automatically formed, through clustering, in the language model when it has not been explicitly designed to model latent topics. To address this question, we design different probe experiments. Using BERT and DistilBERT, we find that the attention framework plays a key role in modelling such word topic clusters. We strongly believe that our work paves way for further research into the relationships between probabilistic topic models and pre-trained language models.","PeriodicalId":126309,"journal":{"name":"European Conference on Information Retrieval","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-01-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133533480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：