Preface

Viglet Turing AI (https://viglet.com/turing) is an open source solution (https://github.com/openturing), which has Semantic Navigation and Chatbot as its main features. You can choose from several NLPs to enrich the data. All content is indexed in Solr as search engine.

1. Architecture

Architecture
Figure 1. Turing AI Architecture

2. NLP

Turing support the followings providers:

2.1. OpenNLP

Apache OpenNLP is a machine learning based toolkit for the processing of natural language text.

2.2. OpenText Content Analytics

It transforms data into insights for better decision-making and information management while freeing up resources and time.

2.3. CoreNLP

CoreNLP is your one stop shop for natural language processing in Java! CoreNLP enables users to derive linguistic annotations for text, including token and sentence boundaries, parts of speech, named entities, numeric and time values, dependency and constituency parses, coreference, sentiment, quote attributions, and relations. CoreNLP currently supports 6 languages: Arabic, Chinese, English, French, German, and Spanish.

2.4. SpaCy

It is a free open-source library for Natural Language Processing in Python. It features NER, POS tagging, dependency parsing, word vectors and more.

Website: https://spacy.io

2.5. Polyglot NLP

Polyglot is a natural language pipeline that supports massive multilingual applications.

3. Documents and OCR

It can read PDFs and Documents and convert to plain text and also it uses OCR to detect text in images and images into documents.

4. Semantic Navigation

4.1. Connectors

Semantic Navigation uses Connectors to index the content from many sources.

4.1.1. Apache Nutch

Plugin for Apache Nutch to index content using crawler.

4.1.2. Database

Command line that uses the same concept as sqoop (https://sqoop.apache.org/), to create complex queries and map attributes to index based on the result.

4.1.3. File System

Command line to index files, extracting text from files such as Word, Excel, PDF, including images, through OCR.

4.1.4. OpenText WEM Listener

OpenText WEM Listener to publish content to Viglet Turing.

4.1.5. Wordpress

Wordpress plugin that allows you to index posts.

4.2. Named Entity Recognition (NER)

With NLP it is possible to detect entities such as:

  • People

  • Places

  • Organizations

  • Money

  • Time

  • Percentage

4.3. Facets

Define attributes that will be used as filters for your navigation, consolidating the total content in your display

4.4. Targeting Rules

Through attributes defined in the contents, it is possible to use them to restrict their display based on the user’s profile.

4.5. SDK Java

Java API (https://github.com/openturing/turing-java-sdk) facilitates the use and access to Viglet Turing AI, without the need for consumer search content with complex queries.

5. Chatbot

Communicate with your client and elaborate complex intents, obtain reports and progressively evolve your interaction.

Its components:

5.1. Agent

Handles conversations with your end users. It is a natural language processing module that understands the nuances of human language

5.2. Intent

An intent categorizes an end user’s intention for taking a conversation shift. For each agent, you define several intents, where your combined intents can handle a complete conversation.

5.3. Actions

The field of action is a simple field of convenience that helps to execute logic in the service.

5.4. Entity

Each intent parameter has a type, called an entity type, that dictates exactly how the data in an end user expression is extracted.

5.5. Training

Defines and corrects intents.

5.6. History

Shows the conversation history and reports.

6. OpenText Blazon Integration

Turing AI detects Entities of OpenText Blazon Documents using OCR and NLP, generating Blazon XML to show the entities into document.

7. Turing AI Console

Turing AI has many components: Search Engine, NLP, Converse (Chatbot), Semantic Navigation

7.1. Login

When access the Turing AI, appear a login page. For default the login/password is admin/admin

Login
Figure 2. Login Page

7.2. Search Engine

7.2.1. Configuration

Search Engine is used by Turing to store and retrieve data of Converse (Chatbot) and Semantic Navigation Sites.

Search Page
Figure 3. Search Engine Page

It is possible create or edit a Search Engine with following attributes:

Table 1. Search Engine Attributes
Attribute Description

Name

Name of Search Engine

Description

Description of Search Engine

Vendor

Select the Vendor of Search Engine. For now it only supports Solr.

Host

Host name where the Search Engine service is installed

Port

Port of Search Engine Service

Language

Language of Search Engine Service.

Enabled

If the Search Engine is enabled.

7.3. Semantic Navigation

7.3.1. Configuration

Semantic Navigation Page
Figure 4. Semantic Navigation Page
Detail Tab

The Detail of Semantic Navigation Site contains the following attributes:

Table 2. Semantic Navitation Site Detail
Attribute Description

Name

Name of Semantic Navigation Site.

Description

Description of Semantic Navigation Site.

Search Engine

Select the Search Engine that was created in Search Engine Section. The Semantic Navigation Site will use this Search Engine to store and retrieve data.

NLP

Select the NLP that was created in NLP Section. THe Semantic Navigation Site will use this NLP to detect entities during indexing.

Thesaurus

If will use Thesaurus.

Language

Language of Semantic Navigation Site.

Core

Name of core of Search Engine where will be stored and retrieved the data.

Fields Tab

Fields Tab contains a table with the following columns: .Semantic Navitation Site Fields Columns

Column Name Description

Type

Type of Field. It can be:

- NER (Named Entity Recognition) used by NLP.

- Seach Engine used by Solr.

Field

Name of Field.

Enabled

If the field is enabled or not.

MLT

If this field will be used in MLT.

Facets

To use this field like a facet (filter)

Highlighting

If this field will show highlighted lines.

NLP

If this field will be processed by NLP to detect Entities (NER) like People, Organization and Place.

When click in Field appear a new page with Field Details with the following attributes:

Table 3. Semantic Navitation Site Fields Detail Attributes
Attribute Description

Name

Name of Field

Description

Description of Field

Type

Type of Field. It can be: INT, LONG, STRING, DATE and BOOL

Multi Valued

If is a array

Facet Name

Name of Label of Facet (Filter) on Search Page.

Facet

To use this field like a facet (filter)

Highlighting

If this field will show highlighted lines.

MLT

If this field will be used in MLT.

Enabled

If the field is enabled.

Required

If the field is required.

Default Value

Case the content is indexed without these field, that is the default value.

NLP

If this field will be processed by NLP to detect Entities (NER) like People, Organization and Place.

Appearance Tab

Contains the following attributes:

Table 4. Semantic Navitation Site Appearance Attributes
Section Attribute Description

Appearance

Number of items per page

Number of items that will appear in search.

Facet

Facet enabled?

If it will be show Facet (Filters) on search.

Number of items per facet

Number of items that will appear in each Facet (Filter).

Highlighting

Highlighting enabled?

Define whether to show highlighted lines.

Pre Tag

HTML Tag that will be used on begin of term. For example: <mark>

Post Tag

HTML Tag that will be used on the end of term. For example: </mark>

MLT

More Like This enabled?

Define whether to show MLT

Default Fields

Title

Field that will be used as title that is defined in Solr schema.xml

Text

Field that will be used as title that is defined in Solr schema.xml

Description

Field that will be used as description that is defined in Solr schema.xml

Date

Field that will be used as date that is defined in Solr schema.xml

Image

Field that will be used as Image URL that is defined in Solr schema.xml

URL

Field that will be used as URL that is defined in Solr schema.xml

7.3.2. Site Page

HTML

In Turing AI Console > Semantic Navigation > <SITE_NAME>, click in Configure button and click Search Page button.

It will open a Search Page that uses the pattern:

GET http://localhost:2700/sn/<SITE_NAME>
JSON

This page requests the Turing Rest API via AJAX. For example, to return all results of Semantic Navigation Site in JSON Format:

GET http://localhost:2700/api/sn/<SITE_NAME>/search?p=1&q=*&sort=relevance
Table 5. Semantic Navigation Rest API Get Attributes
Attribute Required / Optional Description Example

q

Required

Search Query.

q=foo

p

Required

Page Number, first page is 1.

p=1

sort

Required

Sort values: relevance, newest and oldest.

sort=relevance

fq[]

Optional

Query Field. Filter by field, using the following pattern: FIELD: VALUE.

fq[]=title:bar

tr[]

Optional

Targeting Rule. Restrict search based in: FIELD: VALUE.

tr[]=department:foobar

rows

Optional

Number of rows that query will return.

rows=10

8. Customer Case Studies

8.1. Insurance Company

On Intranet of Insurance Company uses OpenText WEM and OpenText Portal integrated with Dynamic Portal Module, a consolidated search was created in Viglet Turing AI, using the connectors: WEM, Database with File System. In this way it was possible to display all the contents and files of the search Intranet, with targeting rules, allowing only to display content that the user has permission. The OpenText Portal accesses Viglet Turing AI Java API, so it was not necessary to create complex queries to return the results.

8.2. Government Company

A set of API Rest was created to make all Government Company content available to partners. All these contents are in OpenText WEM and the WEM connector was used to index the contents on Viglet Turing AI. A Spring Boot application was created with the Rest API set that consumes Turing AI content through the Viglet Turing AI Java API.

8.3. Brazilian University

Brazilian University website was developed using Viglet Shio CMS (https://viglet.com/shio), and all contents are indexed in Viglet Turing AI automatically. This configuration was made in content modeling and the development of the search template was made in Viglet Shio CMS.