Matches in SemOpenAlex for { <https://semopenalex.org/work/W3136996026> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W3136996026 abstract "Topical classification on Lexis Advance uses mostly automatic indexing tools, whether it is machine learning or hand based rules. Even with optimal F measure, these rules can produce a variety of results, going from very relevant to more marginal. With the confidence score project, we want to make sure customers can quickly access the most relevant results, whether they are launching a topic search or setting a topic alert. This will be achieved by stamping within documents, in addition to a topic code, a score indicative of its degree of relevance on a scale of 50 to 99. The new confidence score mechanism is designed to be agnostic, usable by any jurisdiction and for any type of indexing. The system uses Boolean rules as a starting point to generate topic-specific document features that are the input to models to predict the relevance score range for the topic. These generated features include number of words in doc, number of search term matches, position of first and last term match, number of terms in Boolean, ratio of term matches to number of words, and many more. A large number of annotated topic-document pairs were created for training and testing the models to predict the Confidence score ranges. We measured annotator IRR to establish a human level of performance. We experimented with many shallow learning models. Due to several factors we have initially limited our models to those that can be represented as a function inside HPCC in ECL without importing external functions. This limited us to linear regression and decision trees. Many variations of hyper-parameters and sampling strategies were tried in order to achieve a proper balance between precision and recall. Built into the infrastructure, we will also have the ability to force a major reference score for specific LNIs or for documents that contain specific vocabulary. The first step of implementation will be on the HK adaptation, where a threshold will be established so that customers who set up alerts can choose to only receive major references, which has been a pain point in the past. As a second step, the score can be used to sort topic search results based on relevancy. Lastly, the confidence score will be a key component in any project that plans to leverage legal topics, whether it is analytics or matching user queries to topics to boost result relevancy." @default.
- W3136996026 created "2021-03-29" @default.
- W3136996026 creator A5020597558 @default.
- W3136996026 creator A5038164926 @default.
- W3136996026 creator A5048217446 @default.
- W3136996026 creator A5054858528 @default.
- W3136996026 date "2020-11-05" @default.
- W3136996026 modified "2023-09-25" @default.
- W3136996026 title "Topic Classification – Confidence is Key!" @default.
- W3136996026 hasPublicationYear "2020" @default.
- W3136996026 type Work @default.
- W3136996026 sameAs 3136996026 @default.
- W3136996026 citedByCount "0" @default.
- W3136996026 crossrefType "posted-content" @default.
- W3136996026 hasAuthorship W3136996026A5020597558 @default.
- W3136996026 hasAuthorship W3136996026A5038164926 @default.
- W3136996026 hasAuthorship W3136996026A5048217446 @default.
- W3136996026 hasAuthorship W3136996026A5054858528 @default.
- W3136996026 hasConcept C119857082 @default.
- W3136996026 hasConcept C121332964 @default.
- W3136996026 hasConcept C124101348 @default.
- W3136996026 hasConcept C136197465 @default.
- W3136996026 hasConcept C138885662 @default.
- W3136996026 hasConcept C154945302 @default.
- W3136996026 hasConcept C158154518 @default.
- W3136996026 hasConcept C159985019 @default.
- W3136996026 hasConcept C17744445 @default.
- W3136996026 hasConcept C192562407 @default.
- W3136996026 hasConcept C199539241 @default.
- W3136996026 hasConcept C204323151 @default.
- W3136996026 hasConcept C23123220 @default.
- W3136996026 hasConcept C2524010 @default.
- W3136996026 hasConcept C28719098 @default.
- W3136996026 hasConcept C33923547 @default.
- W3136996026 hasConcept C41008148 @default.
- W3136996026 hasConcept C41895202 @default.
- W3136996026 hasConcept C516687495 @default.
- W3136996026 hasConcept C61797465 @default.
- W3136996026 hasConcept C62520636 @default.
- W3136996026 hasConcept C75165309 @default.
- W3136996026 hasConceptScore W3136996026C119857082 @default.
- W3136996026 hasConceptScore W3136996026C121332964 @default.
- W3136996026 hasConceptScore W3136996026C124101348 @default.
- W3136996026 hasConceptScore W3136996026C136197465 @default.
- W3136996026 hasConceptScore W3136996026C138885662 @default.
- W3136996026 hasConceptScore W3136996026C154945302 @default.
- W3136996026 hasConceptScore W3136996026C158154518 @default.
- W3136996026 hasConceptScore W3136996026C159985019 @default.
- W3136996026 hasConceptScore W3136996026C17744445 @default.
- W3136996026 hasConceptScore W3136996026C192562407 @default.
- W3136996026 hasConceptScore W3136996026C199539241 @default.
- W3136996026 hasConceptScore W3136996026C204323151 @default.
- W3136996026 hasConceptScore W3136996026C23123220 @default.
- W3136996026 hasConceptScore W3136996026C2524010 @default.
- W3136996026 hasConceptScore W3136996026C28719098 @default.
- W3136996026 hasConceptScore W3136996026C33923547 @default.
- W3136996026 hasConceptScore W3136996026C41008148 @default.
- W3136996026 hasConceptScore W3136996026C41895202 @default.
- W3136996026 hasConceptScore W3136996026C516687495 @default.
- W3136996026 hasConceptScore W3136996026C61797465 @default.
- W3136996026 hasConceptScore W3136996026C62520636 @default.
- W3136996026 hasConceptScore W3136996026C75165309 @default.
- W3136996026 hasLocation W31369960261 @default.
- W3136996026 hasOpenAccess W3136996026 @default.
- W3136996026 hasPrimaryLocation W31369960261 @default.
- W3136996026 hasRelatedWork W1599651760 @default.
- W3136996026 hasRelatedWork W1975680564 @default.
- W3136996026 hasRelatedWork W2469819362 @default.
- W3136996026 hasRelatedWork W2839451800 @default.
- W3136996026 hasRelatedWork W2888021051 @default.
- W3136996026 hasRelatedWork W2890623459 @default.
- W3136996026 hasRelatedWork W2914769449 @default.
- W3136996026 hasRelatedWork W2949427883 @default.
- W3136996026 hasRelatedWork W2949709209 @default.
- W3136996026 hasRelatedWork W2952609109 @default.
- W3136996026 hasRelatedWork W3042833717 @default.
- W3136996026 hasRelatedWork W3085666889 @default.
- W3136996026 hasRelatedWork W3125183455 @default.
- W3136996026 hasRelatedWork W3126449980 @default.
- W3136996026 hasRelatedWork W3169125662 @default.
- W3136996026 hasRelatedWork W3175609992 @default.
- W3136996026 hasRelatedWork W3212594339 @default.
- W3136996026 hasRelatedWork W783721745 @default.
- W3136996026 hasRelatedWork W2249599227 @default.
- W3136996026 hasRelatedWork W2557556902 @default.
- W3136996026 isParatext "false" @default.
- W3136996026 isRetracted "false" @default.
- W3136996026 magId "3136996026" @default.
- W3136996026 workType "article" @default.