Tip #44 : Using LitSense for Sentence Retrieval from PubMed and PMC

Many thanks to Erica Lake (Outreach Coordinator, NNLM Region 6 ) for this week's tip!

LitSense is a free, web-based tool from NLM that enables simultaneous semantic searching of PubMed and PMC content. LitSense searches at the sentence-level rather than the article-level, allowing for direct retrieval of specific statements. And it searches by best-match and by meaning, retrieving articles containing sentences with an exact match of keywords as well as those with semantic similarity.

Popular uses of this product

• Discover similar findings across different studies to compare and contrast.
• Locate and validate publications for evidence attribution.
• Identify MeSH terms and keyword variations for a PubMed search by reviewing the semantically similar results.
• Retrieve citations containing a phrase used by requestors when discussing their search topic or completing a search request form.
• Non-experts can type in a question as a sentence and retrieve relevant information without having to sift through full-text articles. 

Key Features: 

• The retrieved sentences are presented as “snippets.” Matching terms in each sentence are highlighted in bold.

Snippet from PMID 35845164

• A colored vertical line represents the degree to which the returned sentence is similar to the user query, from orange (high) to green (moderate) to blue (low). 

• Results can be filtered by either Publication Year or Article Section (Methods, Results, etc.).
• The color legend for highlighting biomedical entities, like gene and cell line, can be turned on or off by using the BioConcepts menu in the upper right corner of the results page.
• Clicking on the hyperlinked title or the snippet’s Context button allows users to jump directly to sentence inside the abstract or full-text.
• Clicking on the PMID or PMC number takes the user directly to PubMed or PMC where citations can be saved or exported.

Note:

• By default, LitSense returns relevant sentences that match at least 60% of the terms in a query. When there is not a 60% match result, a 30% match result is returned with the message, “There are no good matches for your query. Showing less related matches.” If there is not a 30% match result, this message is returned: “No sentences share at least 30% of words with your query, or your sentence refers to a publication that was not yet added to our database.” To make some part of your query mandatory, surround it by double quotes. For example, for the query "measles outbreak" vaccination, only documents containing "measles outbreak" will be shown.
• While LitSense searches all of the more than 35 million abstracts in PubMed, it searches just the Text Mining subset of PMC – these are the nearly three million Open Access BioC format articles.
• LitSense is a supported NLM research project and not an official product.

For more information:

Alexis Allot, Qingyu Chen, Sun Kim, Roberto Vera Alvarez, Donald C Comeau, W John Wilbur, Zhiyong Lu, LitSense: making sense of biomedical literature at sentence level, Nucleic Acids Research, Volume 47, Issue W1, 02 July 2019, Pages W594–W599, https://doi.org/10.1093/nar/gkz289