Abstract
In biomedicine, key concepts are often expressed by multiple words (e.g., ‘zinc finger protein’). Previous work has shown treating a sequence of words as a meaningful unit, where applicable, is not only important for human understanding but also beneficial for automatic information seeking. Here we present a collection of PubMed® Phrases that are beneficial for information retrieval and human comprehension. We define these phrases as coherent chunks that are logically connected. To collect the phrase set, we apply the hypergeometric test to detect segments of consecutive terms that are likely to appear together in PubMed. These text segments are then filtered using the BM25 ranking function to ensure that they are beneficial from an information retrieval perspective. Thus, we obtain a set of 705,915 PubMed Phrases. We evaluate the quality of the set by investigating PubMed user click data and manually annotating a sample of 500 randomly selected noun phrases. We also analyze and discuss the usage of these PubMed Phrases in literature search.
Highlights
Background & SummaryUnlike other general domains, the language of biomedicine uses its own terminology to describe scientific discoveries and applications
We examined the composition of the set and found that 84.1% of the phrases are noun phrases
We further noticed from the PubMed user logs that, given documents scored by BM25, users are four times more likely to click on a document containing query terms in the title than on a document that does not
Summary
The language of biomedicine uses its own terminology to describe scientific discoveries and applications. Collocations are restricted to noun/adjective phrases or phrasal verbs, whereas we do not limit phrases grammatically, but rather see them as more flexible entities to be used as building blocks to form longer phrases or sentences Such an interpretation of phrases is better aligned with our goal of using the corpus to analyze queries, as queries may frequently contain incomplete phrases and, in general, are known to differ from traditional forms of written language[4]. To compute and compare the retrieval performance in the absence of a manually annotated gold standard, we use a novel pseudo-relevance judgement technique, which is based on the assumption that the documents containing query terms in the titles are more relevant to the query than the documents that do not[13] Guided by this evaluation, we collect a set of 705,915 multi-word strings that benefit from being interpreted as phrases rather than individual tokens in terms of retrieval performance. Throughout this paper, the term phrase refers to a coherent chunk of words that are frequently used together
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
