Glossary

A piece of documentation within a search strategy.

See: Docs > Strategies > Annotating a Strategy

Acronym for Application Programming Interface: a software intermediary that allows two applications to talk to each other. In the context of Sparque Desk it means the set of API endpoints that make up a service. For a typical search application there may be the following endpoints: an autocomplete, a search, a type-facet, an author-facet and a recommender endpoint. This set works as a whole; the service isn't complete if one of the endpoints in the set is not available. An active API is tied to a specific version of the data: each endpoint will yield information from the same version of the data (all endpoints will upgrade simultaneously).

See: Docs > APIs

A component of an API that forms a touchpoint in the communication with other systems. In Sparque Desk API endpoints are tied to search strategies to expose their functionality to the outside world.

An API endpoint is part of a Sparque request URL:

https://rest.sparque.ai/3.0/{WORKSPACE}/api/{API}/e/{ENDPOINT}/p/{PARAMETER}/{VALUE}/results

In the example below the 'search' endpoint of the 'movies' API is requested to return results for the query 'pulp fiction':

https://rest.sparque.ai/3.0/fliques/api/movies/e/search/p/query/pulp%20fiction/results

See: Docs > APIs > Endpoints

The marker that defines which output connector in a strategy is selected as the output of the strategy as a whole:

See: Docs > Strategies > Setting the Output of a Strategy

A pattern an API Endpoint or a search strategy conforms to. It denotes the parameter(s) it expects and the type of output it returns.

It can be found in the overview of endpoints within an API:

and in the toolbar of the strategy editor:

In this case the endpoint/strategy expects the parameter 'query' of type STRING and returns results of type OBJ, an object from the database.

See: Docs > APIs, Docs > Strategies > Defining the API template

In context of a graph: a triple specifying a Literal value for a property. (not a relation between 2 nodes).

A constituent part of a search strategy. Each block performs an operation on the knowledge graph: filtering, ranking, transforming, combining, matching, etc. By combining blocks operations of a higher complexity can be defined. Blocks are combined by connecting the output connector of one block to the input connector of another block.

Under the hood a block contains a SpinQL snippet that defines how the probabilistic graph database should be queried in order to achieve the desired result.

A set of blocks doing a specific subtask within a search strategy.

Temporary space used for making strategies/APIs execute faster.

The specification of a concept.

A type of database that stores data using a column oriented model. They are very scalable and fast to load and query and therefore suitable for powering search engines.

An immutable instance of a Strategy Document.

Depending on the context it can mean a block->spinql compiler, a spinql->sql compiler or a sql->mal compiler. Usually we mean the spinql->sql compiler, which determines how to execute a search strategy and what to add to the cache to make execution most efficient.

A mental representation of a category of objects in a certain domain.

A small circle at the top of a block (input connector) or at the bottom of a block (output connector) that allows blocks of matching data types to be connected to each other.

The output connector of a Datasets block being connected to the input connector of a Filter by Class block

See: Connecting Building Blocks

Describes how to access external data that needs to be loaded. May be flat files (like csv, xml or json, XLSX), but also allows access to remote data (for example JDBC) or compressed contents.

The description of how data is organized in a database. In the case of a relational database it defines, for example the tables, their fields and the relations between them. See: Knowledge Graph

A collection of data represented as a knowledge graph.

A combination of data sources and their mappings.

A data system that is incompatible or not integrated with other data systems. Data silos arise within organizations when different applications are used in the day-to-day business, that each store data in application-specific proprietary formats. As a consequence data is not adequately shared and remains trapped within a container like grain within a silo.

Data in its raw form. Obtained either through files (like CSV, XML, JSON or XLSX) or through protocols (HTTP, JDBC).

A sprint with a focus on understanding the (search) problems of an organisation and exploring ways to solve them. See: https://sparque.ai/services/design-sprint/, Develop Sprint

A design methodology to solve complex problems. User needs form the starting point of an iterative process in which a solution (product or service) is created. Taking a user-centric approach while creating solutions ensures designers are solving actual problems of users instead of assumed problems of non-users. Design thinking comprises three distinct phases (understand, explore and materialize).

A sprint with a focus on exploring ways to solve known (search) problems and materializing them. See: https://sparque.ai/services/develop-sprint/, Design Sprint

A particular subject area on which an application is intended to apply.

The relation between two objects in a graph. See: Predicate

See: API Endpoint

The instantiation of a class. In other words: an object within the domain such as a document, a fact or a product.

A search task people perform to find material dealing with a particular subject. See: Known-item search

The second phase of a design thinking iteration in which designers ideate about ways to solve (search) problems that were identified in the understand phase and create a prototype based on the most promising solution(s).

A (data) structure consisting of nodes (or entities / items / objects / points / resources / vertices) and edges (or attributes / links / lines / relations) between them.

A database that uses graph structures like nodes, edges and properties to represent and store data. See: tore

The type of concept a user needs to complete a (work) task or to satisfy the curiosity of the mind, independent of the method used to address the need, and regardless of whether the need is satisfied or not.

The computer science field that studies how information objects are obtained from a collection in response to an information need that is expressed by a user. Results are typically ranked on the relevance of the information object to the expressed information need.

A person that is responsible for gathering, managing and disseminating data within organizations. Information Specialists have domain knowledge and are well-aware of the information that is available within an organization and its quality. In addition they have a good understanding of the tasks that users perform with the use of data, of the challenges they encounter in their work and how search might help. Information specialists together with users can best determine which items a search application should return and in what order.

A small circle at the top of a block that indicates what type of data a block can have as input. An input connector can be connected to an output connector of another block if their data types match. Depending on the block these can be one or more output connectors. If an input connector is connected to an output connector this results in data entering the block and being transformed by the operation defined for the block.

The input connector of a Filter by Class block

See: Connector, Connecting Building Blocks

A model that integrates data by storing concepts/entities/resources as linked data triples. As such, Knowledge Graphs do not have a fixed data schema and can be used to link any type of data, tiny or large, internal or external, structured, semi-structured or unstructured together into one rich model of a domain.

A search task people perform to find a particular entity they know exists. See: Exploratory search

Data that is identified by Uniform Resource Identifiers (URIs). By identifying data this way, the meaning of data can be determined by looking up the URI and data from different sources can be integrated based on the URI. See: RDF, Triple, https://www.w3.org/DesignIssues/LinkedData.html

Linked Data which is released under an open license, which does not impede its reuse for free. See: https://www.w3.org/DesignIssues/LinkedData.html

The process of ad-hoc populating the knowledge graph (running the pipelines).

The definition of how to map/interpret data in a pipeline to the knowledge graph constituents.

The third and final phase of a design thinking iteration in which designers let users test the prototype they've built in the explore phase and implement the tested solution to structurally improve the experience of the users.

An object in a graph. See: edge

Prepare a strategy for efficient execution.

The name of the third entity/resource in a triple. See: Subject, Predicate, Property

A small circle at the bottom of a block that indicates what type of data a block can have as output. An output connector can be connected to an input connector of another block if their data types match. An output connector can be connected to multiple input connectors. If an output connector is connected to an input connector this results in data flowing to the connected block.

The output connector of a Datasets block

See: Connector, Connecting Building Blocks

A setting within a block or an input for an API endpoint.

The name of the second entity in a triple denoting the relation between the first and the third entity. See: Subject, Object, property, Relation

A database that stores graph data and a probability estimate for every node and every edge. Essentially a triplestore extended with a probability for every triple. This extension allows for the fusion of selecting and ranking: when selecting structured data from a database, one is presented with certain answers from facts; the probability estimates will be equal to one or zero. When ranking unstructured data, computers will never be certain about the degree of relevance of the result - here, the answer to a query is really a series of probable answers, produced by one ranking algorithm or another. The PGD stores facts with a probability of 1 and ranked results with a probability estimate that expresses their relevance. See: SpinQL

The Resource Description Framework: a data model that specifies how triples can be stored.

See: Triplestore

A connection/edge between two nodes.

A computer language that is used to query a database.

The design process in which design thinking is applied to the creation of search solutions.

A combination of blocks in Sparque Desk. A search strategy results in a desired ranking of the entities in a knowledge graph. Under the hood the constituent SpinQL snippets that make up the blocks define one complex SpinQL query that selects and ranks the desired entities in the probabilistic graph database.

An activity people perform to gather information using information systems. While designing a search application, search designers interview users about their most important search tasks. They then include these in the backlog of the application. See: Seeking Task, Task, Work Task

An activity people perform to gather information using various sources such as people, paper-based documents, information systems, etc. See: Search Task, Task, Work Task

Data that is described by markup. It doesn't conform to the strict structure of a relational database but enforces hierarchies of records and fields within the data. Examples are XML and JSON. See: Structured Data, Unstructured Data

A version of a strategy that was explicitly named and saved to be used by an API endpoint or to be reverted to.

See: Saving a Version of a Strategy

The query language Sparque developed to query the probabilistic graph database (PGD). For all operations on the PGD it has been defined how to compute the resulting probabilities using SpinQL. SpinQL thus combines and propagates probabilities allowing the fusion of selecting and ranking.

A fixed time event (usually one or two weeks) during which a Scrum Team turns abstract ideas into concrete functionality of the highest possible value. See: Design Sprint, Develop Sprint, https://scrumguides.org/scrum-guide.html#the-sprint

Structured Query Language: the computer language that is used for managing data in a relational database.

A Search Strategy that takes the output of another Search Strategy as its input.

See: Stacking Strategies

See: Search Strategy

The implementation of a search strategy.

A parameter that is used within a strategy that can be filled in with a real value on demand.

See: Docs > Strategies > Adding a strategy parameter

Data that is stored in tables (typically in relational databases, spreadsheets, csv’s etc.). It is often quantitative data, such as numbers and dates. Other examples are phone numbers, zip codes and customer names. See: Semi-Structured Data, Unstructured Data

The name of the first entity/resource in a triple. See: Predicate, Object, Property

An activity people perform to accomplish a goal. See: Search Task, Seeking Task, Work Task

See: API Template

An instantiation of one or more strategy parameters. If for example a strategy parameter 'query' is defined, a test could be 'tarantino'. This simulates the case where the user of a movie site would search for 'tarantino'. For a test to be used for a strategy it must match its API Template.

A list of tests.

A constituent part of a knowledge graph of the form subject - predicate - object. One resource (the subject) is connected to another resource (the object) via a relation (the predicate).

(Or RDF store): a purpose-built database for the storage and retrieval of triples.

The first phase of a design thinking iteration in which designers get to know an organization, empathize with the users and define their most pressing problem(s).

Data that is stored in its native format. Even though it may have an internal structure, it's not structured in a predefined way. Unstructured data is often qualitative data, such as customer surveys and interviews. Other examples are text files, multimedia content and e-mail. See: Semi-Structured Data, Structured Data

Acronym for Uniform Resource Identifier: a unique sequence of characters that identifies a resource used by web technologies.

A closed environment within Sparque Desk in which designers can create datasets, pipelines, strategies, APIs, testsets.

An activity people perform to fulfill the responsibility for their work. See: Search Task, Seeking Task, Task

A (relatively small) piece of data (ideally no more than a few megabytes) coming from a container to be processed by the load process. A container specifies the size of the XML Fragment. Often Container's provide mechanisms to influence the size of an XML Fragment.