Skip to content

Documentation

File View

Read one document in detail and review its topic, attributes, entities, relations, and coreference clusters.

Overview

File View is the review screen for one document, also called the evidence reader. The document text fills the reading area on the left. The review panels on the right show what IntQuest extracted and let you correct it: the topic, attributes, entities, relations, and coreference clusters.

  • 1 Header - Back and the file name.
  • 2 Document text - the reading area, with extracted entities highlighted.
  • 3 Entity navigator - finds entities in the document.
  • 4 Topic - the document language and topic.
  • 5 Attributes - the categories that classify entities.
  • 6 Entities - the entities extracted from the document.
  • 7 Relations - the links between two entities.
  • 8 Coreference Clusters - mentions that refer to the same thing.

On screens narrower than about 1,000 pixels, the review panels move below the document text.

Open a document

You can open File View from several places:

  • In Search, click a file name or Open document (eye icon). If you searched with a keyword filter, File View opens with the matching entity, topic, relation, or attribute already highlighted.
  • In Files, click a file name or row.
  • In Stats, click the eye icon in the document list of a chart value.
  • In Text Drilldown, click a document name.

Header and document text

  • 1 Back - returns to the previous page.

  • 2 File name - the name of the open document.

  • Back returns to the previous page. If Esc is the first key you press after the page opens, it also goes back.

  • The file name appears next to Back.

  • The document text is shown on a quiet reading surface. Extracted entities are highlighted in the color of their attribute.

Highlight actions

When you point at or focus a highlighted entity, a small toolbar appears with these buttons. On touch screens, the buttons are always visible.

  • Add to cluster merge: adds the entity to the merge drawer. If the drawer is closed, it opens.
  • Set relation entity: fills the active side of the relation builder. It appears only while the relation builder is open.
  • Locate entity in entities panel: scrolls the Entities panel to the entity.
  • Remove entity: deletes the entity from the document at once, without a confirmation.

Clicking the highlighted text itself doesn’t select or change anything.

Tag new text

  1. In the Attributes panel, select the attribute to tag with. A Tagging as label shows the selected attribute.
  2. In the document text, select the text to tag.

A new highlight can’t overlap an existing one, and a selection can’t start or end with a line break. If it does, IntQuest shows a message and doesn’t create the entity.

Entity navigator

The entity navigator at the top of the review panels finds entities that IntQuest already extracted in this document.

  1. Click Search entities… to expand the navigator.
  2. Type part of an entity’s text and press Enter, or click the search button.
  3. Use the up and down arrows to move between matches. The counter, such as 1/3, shows your position, and the document scrolls to each match.
  4. Click X to close the navigator and clear the search.

  • 1 Search entities… - the search term.
  • 2 Search button - runs the search.
  • 3 Up and down arrows - move to the previous or next match.
  • 4 X - closes the navigator and clears the search.
KeyAction
EnterRun the search, or move to the next match
Shift+EnterMove to the previous match
Up arrowMove to the previous match
Down arrowMove to the next match
EscClose the navigator

The search is case-insensitive. A * at the start or end of the term is ignored.

Topic panel

The Topic panel shows and changes the topic of the document.

  • 1 Language badge - the document language.

  • 2 Topic list - the assigned topic, or Select topic….

  • The language badge, such as Romanian, shows the document language. Topics are language-specific, so the list shows only the topics of this language.

  • The topic list shows the assigned topic and its confidence score. When no topic is assigned, it shows Select topic….

Select a topic

Click the topic list and click a topic. The change is saved at once. A topic that you choose yourself gets a confidence of 100%.

  • 1 Language badge - the list shows only topics of this language.
  • 2 Topic list - click it, or type to search or create a topic.
  • 3 Topics - click a topic to assign it.

Create a topic

  1. Click the topic list and type the new topic name.
  2. Click the Create “…” option at the end of the list.

IntQuest creates the topic for the document’s language and assigns it to the document. Topic names that you create here can contain only letters, including accented letters, and underscores. If the name contains spaces, digits, or other characters, the Create option doesn’t appear. To create topics with spaces or to add seed words, use Administration > Topics.

Important behavior

  • A topic created for Romanian documents doesn’t appear in the list for English documents.
  • If the document’s topic isn’t in the list of its language, it still appears as selected. You can switch to a listed topic at any time.

Attributes panel

The Attributes panel manages the categories, such as PERSON or ORGANIZATION, that classify entities.

  • 1 New attribute - the name of an attribute to create.
  • 2 + (Create attribute) - creates the attribute.
  • 3 Other - an attribute group. Default attributes are under Defaults.
  • 4 Attribute tag - the default checkbox, entity count, name, color, and delete buttons.

Attributes are grouped under Defaults and Other. Each attribute tag shows:

  • a Set as default filter checkbox. Default attributes are shown when no single attribute is selected.
  • the number of entities with that attribute in this document.
  • the attribute name. Click it to select the attribute; click it again to clear the selection. While an attribute is selected, only its entities and relations are shown, and new text that you select is tagged with it.
  • Change color for attribute (palette icon): opens the Change Attribute Color popover with a preview, predefined colors, and a custom color. Click Save to apply the color to the attribute’s highlights.
  • X: deletes the attribute.

To create an attribute, type its name in New attribute and press Enter, or click + (Create attribute).

When the selected attribute has entities, an Export selected attribute button exports those entities from this document.

Warning: deleting an attribute removes the attribute and all entities with that attribute in every document in the system, together with their relations and cluster mentions. The confirmation dialog shows what will be deleted. You can’t undo it.

Entities panel

The Entities panel lists the entities extracted from the document. The line under the heading counts the names and occurrences, for example 12 names · 30 occurrences · confidence per occurrence. Confidence is the model’s certainty for each occurrence.

  • Search entities…: filters the list by entity text after a short delay. The search is case-insensitive.
  • Export all: exports all entities of the document.

Entities with the same text are grouped into one row. Each row shows the entity text, the number of occurrences when there is more than one, the confidence, and these buttons:

  • View documents containing this entity (eye icon): opens Text Drilldown for the entity text.
  • Scroll to entity in text (crosshair icon): scrolls the document to the entity.
  • Manage attributes (tags icon): opens the Entity Attributes popover. Click X on an attribute badge to remove it, or choose one from Add Attribute… to add it.
  • Delete entity (X): deletes the entity after a browser confirmation that lists the relations and cluster mentions that are deleted with it.

Expand N occurrences on a row to see each occurrence with its own confidence, crosshair, tags, and delete buttons.

Clicking an entity row while the relation builder is open fills the active side of the builder. While you edit or build a cluster, the click adds the entity to the cluster.

If AI processing is turned off on your installation, or the document has no extracted entities yet, the list is empty and the counts are 0.

Relations panel

The Relations panel lists links between two entities. Each row reads subject, relation type, and object, for example Compania Dacia — owned_by — Grupul Renault. The percentage at the end is the model’s confidence that the relation is correct.

Each row has these buttons:

  • Eye icon on the subject or object: opens Text Drilldown for that entity’s text.
  • Crosshair icon on the subject or object: scrolls the document to that entity.
  • Delete relation (X): deletes the relation at once, without a confirmation.

Type in Search relations… to filter the list by subject, relation name, or object.

The relation types that the models produce depend on the language. Expand a list to see the default labels.

Default relation labels for Romanian
  • after_a_work_by
  • applies_to_jurisdiction
  • architect
  • author:screenwriter
  • based_on
  • basin_country
  • capital
  • characters
  • child_astronomical_body
  • competition_class
  • conflict
  • constellation
  • contains_administrative_territorial_entity
  • continent
  • creator
  • creator:author
  • creator:composer
  • creator:developer
  • creator:director
  • creator:founded_by
  • crosses
  • designated_as_terrorist_by
  • diplomatic_relation
  • discoverer_or_inventor
  • distributor
  • employer
  • ethnic_group
  • feast_day
  • field_of_work
  • followed_by
  • follows
  • genre
  • has_effect
  • has_part
  • has_part:contains_settlement
  • has_parts_of_the_class
  • has_quality
  • head_of_government
  • headquarters_location
  • heritage_designation
  • historical_region
  • indigenous_to
  • industry
  • influenced_by
  • instrument
  • language_of_work_or_name
  • language_of_work_or_name:original_language_of_film_or_TV_show
  • language_used
  • languages_spoken_written_or_signed
  • legislative_body
  • licensed_to_broadcast_to
  • located_in_or_next_to_body_of_water
  • located_in_the_administrative_territorial_entity
  • located_in_the_administrative_territorial_entity:capital_of
  • located_in_the_administrative_territorial_entity:country
  • located_in_the_administrative_territorial_entity:country_of_citizenship
  • located_in_the_administrative_territorial_entity:country_of_origin
  • located_on_terrain_feature
  • location
  • location:located_in_present-day_administrative_territorial_entity
  • location:residence
  • location:work_location
  • location_of_formation
  • lowest_point
  • manufacturer
  • member_of
  • member_of:family
  • member_of_political_party
  • military_branch
  • military_rank
  • most_populous_urban_area
  • mother
  • mountain_range
  • mouth_of_the_watercourse
  • movement
  • named_after
  • native_language
  • no_relation
  • nominated_for
  • notable_work
  • occupant
  • occupation
  • official_language
  • operating_system
  • operator
  • original_network
  • owned_by
  • owned_by:parent_organization
  • owner_of
  • owner_of:subsidiary
  • parent_astronomical_body
  • part_of
  • participant
  • participant_of
  • participant_of:league
  • participating_team
  • partner
  • partner_in_business_or_sport
  • performer
  • performer:cast_member
  • place_of_birth
  • place_of_death
  • place_served_by_transport_hub
  • platform
  • position_held
  • position_played_on_team/speciality
  • present_in_work
  • product_or_material_produced
  • publisher
  • record_label
  • relative
  • relative:child
  • relative:father
  • relative:sibling
  • relative:spouse
  • religion
  • replaced_by
  • replaces
  • separated_from
  • shares_border_with
  • significant_event
  • sport
  • sports_season_of_league_or_competition
  • studied_by
  • successful_candidate
  • taxon_rank
  • territory_claimed_by
  • top-level_Internet_domain
  • tributary
  • twinned_administrative_body
  • voice_type
  • winner
Default relation labels for English
  • no_relation
  • org:alternate_names
  • org:city_of_headquarters
  • org:country_of_headquarters
  • org:dissolved
  • org:founded
  • org:founded_by
  • org:member_of
  • org:members
  • org:number_of_employees/members
  • org:parents
  • org:political/religious_affiliation
  • org:shareholders
  • org:stateorprovince_of_headquarters
  • org:subsidiaries
  • org:top_members/employees
  • org:website
  • per:age
  • per:alternate_names
  • per:cause_of_death
  • per:charges
  • per:children
  • per:cities_of_residence
  • per:city_of_birth
  • per:city_of_death
  • per:countries_of_residence
  • per:country_of_birth
  • per:country_of_death
  • per:date_of_birth
  • per:date_of_death
  • per:employee_of
  • per:origin
  • per:other_family
  • per:parents
  • per:religion
  • per:schools_attended
  • per:siblings
  • per:spouse
  • per:stateorprovince_of_birth
  • per:stateorprovince_of_death
  • per:stateorprovinces_of_residence
  • per:title
Default relation labels for Hungarian, Russian, and French
  • per:employee_of(e1,e2)
  • no_relation
  • per:employee_of(e2,e1)
  • org:city_of_headquarters(e1,e2)
  • per:countries_of_residence(e1,e2)
  • per:countries_of_residence(e2,e1)
  • org:country_of_headquarters(e1,e2)
  • org:members(e2,e1)
  • per:stateorprovinces_of_residence(e1,e2)
  • org:subsidiaries(e2,e1)
  • per:cities_of_residence(e1,e2)
  • per:title(e2,e1)
  • org:subsidiaries(e1,e2)
  • org:stateorprovince_of_headquarters(e1,e2)
  • org:country_of_headquarters(e2,e1)
  • org:top_members/employees(e2,e1)
  • org:city_of_headquarters(e2,e1)
  • org:alternate_names(e1,e2)
  • org:top_members/employees(e1,e2)
  • org:founded(e1,e2)
  • per:country_of_birth(e1,e2)
  • per:cities_of_residence(e2,e1)
  • org:alternate_names(e2,e1)
  • per:title(e1,e2)
  • per:stateorprovinces_of_residence(e2,e1)
  • org:members(e1,e2)
  • org:founded_by(e1,e2)
  • org:founded_by(e2,e1)
  • org:founded(e2,e1)
  • per:spouse(e1,e2)
  • per:origin(e2,e1)
  • per:origin(e1,e2)
  • org:stateorprovince_of_headquarters(e2,e1)
  • per:spouse(e2,e1)
  • per:country_of_birth(e2,e1)
  • per:alternate_names(e1,e2)
  • per:alternate_names(e2,e1)

Add a relation

  • 1 − - closes the relation builder.
  • 2 New - marks a relation that isn’t saved yet.
  • 3 Left Entity - the subject of the relation.
  • 4 Relation… - the relation name.
  • 5 Right Entity - the object of the relation.
  • 6 Reset and Save - clear the builder, or save the relation.
  1. Click + (Add new relation) next to Search relations…. The relation builder opens, marked New, and the button changes to −.
  2. Click Left Entity to make it active, then fill it with an entity: click Set relation entity on a highlight in the text, or click an entity row in the Entities panel.
  3. Click Right Entity and fill it the same way.
  4. In Relation…, type the relation name, for example works_for.
  5. Press Enter, or click the save button (disk icon).

If a part is missing, it is outlined in red and the message Incomplete relation appears. Click the reset button (circular arrow icon) to clear the builder. Clicking the active side again makes it inactive.

Coreference Clusters panel

A coreference cluster groups the mentions in a document that refer to the same real-world thing, for example Ion Popescu, Popescu, and he.

  • 1 Search mentions… - filters the clusters by mention text.
  • 2 View clusters in larger dialog - opens the clusters in a dialog.
  • 3 Cluster list - the cluster cards, or No coreference clusters found.

Clusters load when the document opens, using your signed-in session. If AI processing is turned off on your installation, or the document has no clusters, the panel shows No coreference clusters found.

Panel controls

  • Search mentions…: shows only the clusters with a mention that contains the text. The filter applies as you type.
  • Merge clusters (object-group icon): opens the merge drawer. It appears when the document has two or more clusters.
  • View clusters in larger dialog (expand icon): opens all clusters in the Coreference Clusters dialog. Press Esc or click outside to close it.

Cluster cards

Each card is titled Cluster N with the number of mentions, and shows:

  • the cluster confidence: how confident the model is that the mentions belong together.
  • a row for each mention with its text, an eye button that opens Text Drilldown, a crosshair button that scrolls to it, and a bar with the mention confidence: how confident the model is that the highlighted text is a mention.
  • Previous entity and Next entity buttons that step through the mentions in the document.
  • Edit cluster (pen icon) and Delete cluster (trash icon). Deleting asks for confirmation.

Edit a cluster

  1. Click Edit cluster.
  2. To remove a mention, click its X (Remove from cluster). A cluster must keep at least one mention.
  3. To add a mention, click its row in the Entities panel.
  4. Click the save button (Save cluster changes), or click X (Cancel cluster editing) to discard the changes.

Merge and group clusters

The merge drawer combines clusters, or groups standalone entities into a new cluster.

  1. On each highlight that you want to combine, click Add to cluster merge. The Merge Clusters drawer lists the Selected Items. Entities that aren’t in any cluster are marked (standalone).
  2. Optional: click Remove from merge to drop an item.
  3. Click the action button. It appears when at least two items are selected:
    • Merge N Items combines the selected clusters into the one with the lowest number. Standalone entities join that cluster.
    • Group N Entities into Cluster creates a new cluster when only standalone entities are selected. This is the way to create a new cluster.

Click Cancel to close the drawer without changes.

Important behavior

  • Deleting an entity also removes its mentions from clusters. A cluster left without mentions is deleted.
  • Cluster numbers in the merge drawer can differ from the numbers in the panel.

Notes

  • Creating, deleting, and regrouping items changes the review data of the document for every user. Read confirmation messages carefully. Some deletions, such as removing an entity from the text or deleting a relation, happen without a confirmation.
  • If few highlights appear, check whether an attribute is selected, or which attributes are marked as defaults.
  • Use Text Drilldown to find every document that contains some text, not only this one.
  • The entity and relation filters wait about 300 milliseconds after you stop typing. The cluster mention filter applies at once.