Advanced Search and Discovery.

Natural Language search in many languages

NLP can be both leveraged at search and indexing time, further enhancing the employee’s search experience.

• Automatic detection of query languages (more than 135 detected languages)
• Real part-of-speech tagging at the highest level within many languages such as English, French, German, Arabic, Chinese (simplified), Chinese (Traditional), Korean, Danish, Spanish, Finnish, Greek, Italian, Japanese, Dutch, Polish, Portuguese, Romanian, Russian, Swedish, Thai, Norwegian
• Depending on the languages, several speech technologies must apply, such as transliteration (Japanese), compound word splitting (German, for instance), model-based disambiguations, etc.
• Semantic analysis
• Query expansion with native variants (strict synonyms, verb/noun equivalences, US/UK variants, etc.)
• Query expansion with one-way or two-ways synonyms, to be enriched with custom dictionaries
• Rewriting rules (while searching this, then instead search that)

Auto-Correct

Sinequa automatically detects the language used in the query and performs a syntactic analysis of the text for automatic spell checking and phonetic correction of the query. It also includes query auto-completion, query suggestion and similar document suggestion. The behavior of these features can be configured and tailored to specific applications, including the use of custom dictionaries as desired.

• Spelling correction at time search performed so only relevant suggestions used, and options for correcting query terms based on frequency thresholds and “Did you mean” suggestions
• Phonetic search corrects your query with advanced phonetic identification capabilities

Query intent detection

Query intent detection is the art of providing the right information to the right user in its particular context.

Sinequa’s insight engine finds and presents the right information to make it directly actionable instead of just searching for the most relevant document according to pure information retrieval techniques. Two technologies are combined here to make this happen:

• Deep-Learning (TensorFlow) and pre-trained multiple language models using BERT
• Training samples expanded through NLP and custom dictionaries, required to provide the extra context needed by the insight engine according to the situation and the query type

Fielded-search

An intuitive interface that employees can use to handle complex queries, including inline metadata search.

While complex advanced search forms are less and less in use nowadays, sophisticated applications may be required to filter the answers by additional metadata criteria. To avoid overcomplicating the search interface, an employee may use the simple search box to type complex queries that include multiple criteria applying to metadata. The syntax is straight forward assuming the employee is familiar with the "from:" "to:" or "site:" already in use with major email systems or public search engines.

Advanced search

Run complex searches by filling up customizable advanced search forms.

An advanced search is easily set up by adding as many search criteria as needed by the use-case and combined at will. Sources, formats, dates range, authors, and any custom metadata can be added within advanced forms, along with rich and intuitive features like multi-selection, sliders, exclusions, etc.

Auto-suggest

Make your search more intuitive and contextually relevant, thanks to the "Auto-Suggest" functionality.

• Autosuggest "as you type" lets employees select the most relevant topics to the existing corpus
• Suggestions are selected from multiple content types like lexicons, index titles, index terms, extracted named entities (people, companies, locations, etc.), most frequent queries, and many more resources
• Suggestions are respecting the access rights, so no suggested term can reveal information extracted for content the employee is not permitted to read.