Unstructured Text Classification
Turning digital data into operational intelligence. Organising and annotating any short or long form text with readable, consistent topic categories in a knowledge hierarchy. Customers manipulate these topics to perform behavioral analysis, content recommendation, ad targeting, document retrieval, data discovery, request routing, and more.
- Very Large Multi-class classification (over 500,000 nodes)
- Dedicated functions for processing search, social, documents, and HTML
- Syntactic parsing, named entity recognition, part of speech tagging
- Sentiment scoring
- High throughput (over 50,000 documents per second)
- Multi-language support (English, plus 6 other languages and growing)
- Hybrid Bi-LSTM and ConvNet Machine Learning Architecture
- Trained from a corpus of more than 3 billion ngrams
- Understand user behavior based on content consumed or created
- Match ads, inventory, and users for more effective marketing
- Keep advertising opportunities brand safe
- Segment market research participants by demonstrated or declared interest
- Semantically query and retrieve documents by topic labels
- Analyse open-ended customer survey responses
- Understanding business website content to identify B2B opportunities
- Categorise content in videos, based on transcripts
- Automatically label data to quickly train supervised ML models
- Recommend content based on ontological graph connections
£0.00095 to £0.0015 per unit per month
- Free trial available
3 6 2 2 6 8 2 5 4 1 7 7 8 8 7
0203 289 5125
|Software add-on or extension||Yes, but can also be used as a standalone service|
|What software services is the service an extension to||Data collection; ML (training data); ad serving & targeting; end-user applications|
|Cloud deployment model||
|Email or online ticketing support||Email or online ticketing|
|Support response times||Within 12 hours, M-F. Within 48 hours, Sat & Sun.|
|User can manage status and priority of support tickets||No|
|Web chat support||No|
Support provided at no extra cost. Issues/tickets initially conducted via email. Issues are escalated to phone / virtual support as needed. No dedicated support resources.
Dedicated resource for service customization can be purchased for an additional cost (amount negotiated based on project requirements).
|Support available to third parties||Yes|
Onboarding and offboarding
Users are issued credentials and authentication to access the API services, and provided with documentation on the functions available.
Company resources are available to provide recommendations on best practices, or troubleshoot technical issues, as needed. Client libraries for interacting with the API services are available in Python and PHP.
To familiarize new clients with the data format and output, eContext can process initial datasets internally and provide text or JSON output of results.
|End-of-contract data extraction||N/A. No user data is held by the eContext service.|
|End-of-contract process||Access to the services are suspended. The customer is able to store and continue to use any data retrieved from the service during the contract period.|
Using the service
|Web browser interface||No|
|Application to install||No|
|Designed for use on mobile devices||No|
|What users can and can't do using the API||
All core services are offered as endpoints with the API suite. Customers POST all data they wish to classify and annotate through calls to the API, and receive back the classification, sentiment, entity, and NLP data in the API response. Customers can also check their & monitor their usage, create & invoke custom classification rules, run language identification, and other tasks.
Company sets up and provides authentication for customers, issuing API credentials prior to service commencement. Any limitations set on the customers account can only be changed by the Company. Additional API credentials are granted by the Company.
|API documentation formats||HTML|
|API sandbox or test environment||No|
|Description of customisation||
Users can create custom classification hierarchies and rules through the API, which can be invoked to classify content they submit.
Users can request, via a client services relationship, extensions and additions to the core topic hierarchy -- for example, to recognise topics important to a particular consumer group or region; or to recognise topics unique to a certain field or discipline (for example, Pharmaceuticals, or Real Estate Law).
Users can request, via a client services relationship, to receive not only eContext topic classifications but also the equivalent topic in the users current/proprietary topic hierarchy or category system. This is called an "overlay". For example, eContext offers standard overlays to hierarchies from the Internet Advertising Bureau, the North American Industry Classification System (NAICS), and ad targeting systems of Twitter and Facebook.
|Independence of resources||
Services are monitored for usage and auto-scaled when available resources cross a determined threshold.
eContext services are horizontally distributed, so each part of the service is always isolated. Each resource is monitored and can be automatically removed from the cluster if its latency, load, or other performance characteristic is repeatedly out-of-bounds. In these cases, the particular requests being processed by that resource may take longer to return results, or may fail; however, subsequent requests would be routed to healthy resources and proceed normally.
|Service usage metrics||Yes|
Number of requests processed per API endpoint, per period (day, month, year, all-time)
Number of characters processed per API endpoint, per period (day, month, year, all-time)
|Reporting types||API access|
|Supplier type||Not a reseller|
|Staff security clearance||Other security clearance|
|Government security clearance||Up to Baseline Personnel Security Standard (BPSS)|
|Knowledge of data storage and processing locations||Yes|
|Data storage and processing locations||United Kingdom|
|User control over data storage and processing locations||No|
|Datacentre security standards||Managed by a third party|
|Penetration testing frequency||Less than once a year|
|Penetration testing approach||In-house|
|Protecting data at rest||Other|
|Other data at rest protection approach||Information available upon request|
|Data sanitisation process||No|
|Equipment disposal approach||In-house destruction process|
Data importing and exporting
|Data export approach||N/A. No user data is held by the service. All operations are conducted in a RESTful interface, and responses from the service returned to the user for consumption and storage. Results from the processing of user requests are deleted once returned to the user.|
|Data export formats||Other|
|Other data export formats||JSON|
|Data import formats||Other|
|Other data import formats||JSON|
|Data protection between buyer and supplier networks||
|Data protection within supplier network||IPsec or TLS VPN gateway|
Availability and resilience
EContext will use commercially reasonable efforts to make the eContext API available 99.95% of the time. In the event eContext does not meet the goal of 99.95% API availability in any calendar month (“Monthly Uptime Percentage”), Client will be eligible to receive a financial credit attributed to the pro-rata downtime.
“Monthly Uptime Percentage" is calculated by subtracting from 100% the percentage of continuous 5-minute periods during the Service Month in which the eContext API was in a state of “Unavailable Time” as identified by Monitoring Service.
|Approach to resilience||Information available upon request.|
|Outage reporting||Email alerts, and error codes in API responses|
Identity and authentication
|User authentication needed||Yes|
|User authentication||Username or password|
|Access restrictions in management interfaces and support channels||N/A|
|Access restriction testing frequency||At least every 6 months|
|Management access authentication||
Audit information for users
|Access to user activity audit information||Users contact the support team to get audit information|
|How long user audit data is stored for||User-defined|
|Access to supplier activity audit information||No audit information available|
|How long system logs are stored for||User-defined|
Standards and certifications
|ISO/IEC 27001 certification||No|
|ISO 28000:2007 certification||No|
|CSA STAR certification||No|
|Other security certifications||No|
|Named board-level person responsible for service security||No|
|Security governance certified||No|
|Security governance approach||Information available upon request.|
|Information security policies and processes||Information available upon request.|
|Configuration and change management standard||Supplier-defined controls|
|Configuration and change management approach||
Senior engineering leadership vets technical details, and the change is put to consensus where possible negative consequences are raised.
Developers and product leads to implement the change in their local dev environments, then work with the System Administrator to push changes into a testing environment for review. Changes are then moved to a staging environment and tested under unscripted, client-style usage.
New instances including the change are launched in the background of the eContext VPC, pass automated and manual testing and UAT, and then progressively added into the resource pool, removing older versions, until the change is complete.
|Vulnerability management type||Undisclosed|
|Vulnerability management approach||
Maintaining and regularly updating strict access control lists, firewalls, IP and Port Whitelists; applying security updates to all servers every two weeks; reviewing the lists of internal and external users that have access to data once a month, and removing users that no longer require access; and conducting monthly and annual risk compliance audits.
Threat agents are logged, recorded, and added to blacklists and monitoring to safeguard against future threats. Our incident response procedure is reviewed monthly, as well as any time any step of the process is updated.
|Protective monitoring type||Undisclosed|
|Protective monitoring approach||
Threats are dealt with by monitoring hardware for changes in files, processes, or resource consumption, with notifications delivered to Administrators for events outside expected thresholds.
If an event is determined to be an incident, Countermeasures deployed may include, but are not limited to:
(1) Notifying users and stakeholders, internal and external
(2) Removing any rogue code from the server
(3) Updating firewall rules, IP restrictions, and other access controls
(4) Taking compromised server offline until the incident is resolved
(5) Performing a review to document the incident and share details
Most incidents are resolved with 12 hours.
|Incident management type||Undisclosed|
|Incident management approach||
Users report incidents through email/support ticketing, but most incidents are recognized by internal monitoring first.
Incident reports are authored by the company and distributed to affected clients by email.
|Approach to secure software development best practice||Supplier-defined process|
Public sector networks
|Connection to public sector networks||No|
|Price||£0.00095 to £0.0015 per unit per month|
|Discount for educational organisations||No|
|Free trial available||Yes|
|Description of free trial||A capped number of interactions against the service. May also limit the level of detail in the classification response, or the number of possible classifications returned per transaction.|