This section contains functions that are focused on document-level functionality.
The Demo App displays the frequency of unique terms, bigrams, and trigrams from the document. Additional items in the underlying SalienceDocumentDetails structure such as number of sentences and words from the document are not displayed.
Displays the phrase-based and model-based sentiment results. Phrase-based results include the sentiment phrases that contributed to the overall document sentiment score. For each phrase, the sentiment weight for the phrase is provided in the Score column, as well as whether or not the phrase is contained in the LSF (Type=0) or an HSD (Type=1). If the phrase has been negated or intensified (or both), this is indicated in the Modification column.
The Options button displays the Document methods tab of the [Options dialog], where the HSD file to use can be set in the Document sentiment section.
Displays a summary of the document. The length (in number of sentences) of the summary can be modified by clicking the Options button and adjusting the parameter in the Document summary section of the [Options dialog] shown.
Extracts themes from the document. Theme results are ranked by the Score column, which indicates the relative strength of the theme. The Type column will display a 1 if the theme is a "meta-theme", a theme which occurs as a substring of multiple other themes. Phrase-based theme Sentiment and the associated sentiment Evidence is also provided for theme results.
Query-defined topics apply labels to document content based on the labels and associated queries in the current query topic file. By default, the example query topic file provided in the default data directory (
data/salience/tags/sampletags.dat). An Options button is provided to access the [Options dialog] and browse for a different query topic file in the Document topics section.
Topic results display the Topic label, with the number of query terms for the topic that hit within document content in the Hits column. Topic-level Sentiment is provided, based on phrase-based sentiment analysis of content within the document where query hits are found. The Score column is not used for query topic results.
Concept-defined topics apply labels to document content based conceptual matches to the concepts listed in the current concept topic file. By default, the example concept topic file provided in the default data directory (
data/salience/concepts/example.dat). An Options button is provided to access the [Options dialog] and browse for a different concept topic file in the Document topics section.
As with query-defined topic results, the Topic label is displayed. The Score column provides a measure of the match of document content to the concept. Topic-level Sentiment is provided, based on phrase-based sentiment analysis of content within the document where conceptually related terms are found. The Hits column is not used for concept-defined topic results.
This feature was introduced in Salience 5.1.1, and displays the categories and sub-categories identified for the document content. Expanding a category result shows child categories related to the match.
Two sections contain functions that are focused on entity-level functionality. Named Entity Functions return results that are based on the model-based entity extraction provided in Salience. User-Defined Entity Functions require the use of a User Entity List.
Named Entities/User-Defined Entities
Named entity results display the entities extracted from the content via the entity extraction model, augmented by any user customizations. An Options button is provided to access the Entity options within the [Options dialog].
Entity results are ordered by the Count column, which is the number of mentions within the content for that entity. For each entity result, the entity Label is displayed. Entity-level Sentiment based on phrase-based sentiment is shown as well as the sentiment Evidence.
Double-clicking an entity result displays a dialog with further entity details.
Results for user-defined entities are similar to those for named entities, with the exception that the entities are extracted based on the user entity list as specified in the [Options dialog].
Entity relationships are displayed from the content. These are connections between multiple extracted named entities (or user-defined entities as applicable). For each, the Type of relationship is specified, as well as the entities that participate in the relationship.
Entity opinions are displayed from the content. These are opinions associated named entities (or user-defined entities as applicable). For each, the Speaker of the opinion is identified, as well as the Topic of the opinion and Sentiment detected.
The functions in this section return the provided content with markup of specific characteristics.
This markup of the provided content colors each token based on its part-of-speech. The legend at the top of the panel indicates the general part-of-speech class signified by each color. Hovering over a specific token will display the precise part-of-speech tag for that token.
Named Entity Markup
This markup of the provided content colors the extracted named entities. The legend at the top of the panel indicates the entity type signified by each color.
This markup of the provided content highlights the sentiment-bearing phrases within the content. The legend at the top of the panel indicates the sentiment orientation used within the markup.
Entity Opinion Markup
This markup of the provided content highlights extracted entity opinions within the content. The legend at the top of the panel provides information about the colors used.
Get Chunk Tagged Text
This markup of the provided content displays the chunks identified by Salience within the content. Part-of-speech tags for each token are appended to the token. Hovering over a chunk will display the chunk type (ie. noun phrase, verb phrase, etc.)
Updated almost 2 years ago