Domains

The analysis works best when the appropriate domain is specified via the domain field of the Request object.

The domains can be sorted in two basic genres:

  • Media - suitable for analysis of media and news articles.

  • Voice of the Customer - suitable for customer reviews, but also customer support texts (e.g., support emails).

In addition, there are

  • a generic, genre and domain agnostic analysis

  • private custom domains - these domains are tuned for a particular customer according to their needs.

Media Domains

Media domains are suitable for analysis of media and news articles. Clean spelling and grammar is expected.

Available domains: media, news, sport, tabloid, and tech. The media domain yields generally acceptable results, but choosing a more specific domain improves the quality of results, including (1) entity disambiguation (e.g. Barcelona is considered more likely to mean FC Barcelona than the city itself); (2) tag choice.

Supported analyses and languages:

Czech

Dutch

French

German

English

Slovak

Spanish

Named Entities

Numeric Entities

Entity linking

Semantic tags

In addition, there is an alpha support for Polish.

Voice of the Customer Domains

The VoC domains are suitable for customer reviews, but also customer support texts (e.g., support emails).

If a domain is unsupported for a given language (e.g., voc-hospitality for Slovak), a standard generic domain (see below) is used.

Named entities are limited to selected relevant items, but we can create a custom domain that recognizes the same named entities as media domains (see above).

Supported domains:

Domain

Domain parameter

Languages

Entities

Sentiment

Named

General

Domain specific

Numeric

GKB linking

General

voc

en, cs, sk

selected

Restaurants

voc-hospitality

en, cs, de

selected

food, restaurants

Banking

voc-banking

en, cs, sk

selected

banking, insurance

Retail

ask us

Transporation

ask us

Generic Domain

A generic, genre and domain agnostic analysis that is used when no domain is specified. Similar to Media domain, but entities are not linked to the knowledge base.

Notes:

  • Named entities are names of people, organizations, locations and products.

  • General custom entities are mostly standardized keyphrases (including names of sports, common political and economic terms etc.).