Intent exists before words. A child points, a person types, and both want to be understood.

Communicative intent is the goal-directed use of verbal and non-verbal signals to convey something to someone. In web search, it is the goal behind a query. It is older than language and larger than any single signal.

Intentional communication is deliberate: a person uses a gesture, a look or a word to affect someone else. The signal is directed at a listener and adjusted when it fails. A query retyped until it works is intentional in this sense.

Non-intentional communication consists of signals produced without a deliberate aim, such as an infant crying or fussing. Others read meaning into them. Much of what people reveal online, from hesitation to abandoned searches, is of this kind.

Gestures are movements that express needs and intentions, before and alongside speech. Children point before they speak. The pointing finger is the first query.

If the intent to communicate began with words, these gestures would have to wait for language. All four appear before the first birthday.

Eye gaze signals where attention is directed and can make a request, as when a child looks from an object to a parent. It is one of the earliest communicative signals. It shows intent without a single word.

Facial expressions carry emotion and attitude that accompany or replace words. They tell the listener how a message is meant. Text strips them away.

Verbalisations are spoken words and vocal sounds used to deliver a message. They make intent explicit. They are also only one channel among several.

Many autistic children communicate intent differently, using gestures, eye gaze and spoken words in ways that others may not recognise. The intent is there, and its expression takes other forms. Recognising it requires attention to more than words.

A speech-language pathologist assesses and supports people who have difficulty expressing or understanding communicative intent. The work starts from how a person communicates now. It treats every signal as potentially meaningful.

The Communication Matrix, developed by Charity Rowland, is an assessment tool that maps how a person communicates across seven levels, from pre-intentional behaviour to language. It records every form of communication, including those without words. It shows intent developing step by step.

Also discussed at https://www.communicationmatrix.org/

Search queries are the textual input people give search engines, the written trace of an intent. They are short and stripped of context. Search intent is read from them.

Intent detection is the automatic classification of a user's input into predefined goal categories. Dialogue systems and search assistants rely on it to decide what to do. It works only with the categories it was given.

Andrei Broder's taxonomy from 2002 sorts web search goals into informational, navigational and transactional. It remains the standard frame for query intent. It describes queries, and the person behind the query is larger than the three classes.

Speech acts are utterances understood as actions such as asking or requesting. Communicative intent is the goal those acts serve. The act is how the intent becomes visible.

Dialogue acts classify the function of each utterance within a conversation, such as question, answer or acknowledgement. They are the computational counterpart of speech acts. Systems learn to recognise them from annotated dialogues.

A capsule neural network groups neurons into capsules that preserve relationships between features, an architecture proposed by Geoffrey Hinton's group. It has been applied to intent classification in dialogue. It is one of many models that classify text into known goals.

Conversational implicature is meaning conveyed beyond the literal words, which the listener infers. Much communicative intent travels this way. A classifier that reads only the words misses the implied goal.

Paralinguistic cues are features of speech beyond the words, such as tone, pace and pauses. They change what an utterance means. Voice search carries them, and a text transcript loses them.

Sentiment polarity measures whether a text is positive or negative. It says little about what a person wants. An angry query and a calm one can seek the same thing.

Constructive communication rewriting transforms a harmful or hostile utterance into a respectful one while keeping its underlying intent. It separates what someone wants from how they said it. The intent survives the rewording.

What intent needs beyond classification, and the people whose intent is hardest to read. The usual treatment describes signals and classifiers. It leaves out shared attention, intents that fall outside a system's categories, the details a request contains, mental state attribution, the whole range of language impairments, and the standards that make interfaces accessible. For search, that missing part is the difference between assigning a query to a class and understanding what the person wants.

Joint attention is the shared focus of two people on the same object, each aware that the other attends to it. Infants develop it in the first year, before they speak. Communication is built on it, and a search interface has none.

Out-of-scope intent detection recognises when an input matches none of the categories a system knows. Without it, a system forces every request into the nearest known class. Many real needs belong to no predefined class.

Also discussed at https://arxiv.org/abs/1909.02027

If a classifier that sorts known requests well also recognised requests outside its list, both columns would be high. The second column stays far below the first for every system.

Slot filling extracts the specific details of a request, such as a date, a place or a product, alongside its intent. Booking a flight to Lisbon on Friday has one intent and several slots. Intent without slots cannot be acted on.

Also discussed at https://cloud.google.com/dialogflow/es/docs/intents-actions-parameters

Described in a Google patent: https://patents.google.com/patent/US11347801B2/en

If naming the intent meant understanding the request, the whole-frame row would match the intent row. It sits 11.4 points lower on ATIS and 21.5 points lower on Snips.

Social communication disorder is a diagnosis for persistent difficulties with the social use of language, without the restricted and repetitive behaviours of autism. It was introduced in DSM-5 in 2013. It shows that pragmatic difficulty exists on its own.

Augmentative and alternative communication covers methods and devices that supplement or replace speech, from picture boards to speech-generating devices. They give intent a channel when speech is not available. Every interface is a kind of alternative communication.

Also discussed at https://www.asha.org/public/speech/disorders/aac/

Pragmatic competence is the ability to use and interpret language appropriately in context. It varies between people. Differences in understanding are often differences in this ability, beyond noise in the signal.

An illocutionary force indicating device is a linguistic marker of what an utterance does, such as mood, word order or a question mark. It signals intent in the structure of the words. Detection that relies only on vocabulary ignores it.

Zero-shot intent classification assigns intents that a model has not seen in training, using descriptions of the categories. It extends detection to new domains without labelled data. It depends on how well the categories are named.

Also discussed at https://huggingface.co/tasks/zero-shot-classification

A task-oriented dialogue system helps a user complete a goal, such as booking or troubleshooting, across several turns. It tracks intent, slots and progress. A single classified query is only its first step.

Developmental language disorder is a persistent difficulty with learning language that is not explained by another condition. It affects a notable share of children. Their communicative intent is intact and its expression is harder.

Also discussed at https://www.nidcd.nih.gov/health/developmental-language-disorder

Prosodic features are measurable properties of speech such as pitch contour, stress and rhythm. They mark questions, emphasis and emotion. Spoken intent is carried as much by prosody as by words.

Theory of mind is the ability to attribute beliefs, desires and intentions to others. Understanding communicative intent depends on it. Many misunderstandings are failures of mental state inference, beyond failures of encoding.

A deictic gesture points to something in the shared environment, such as pointing at an object. It resolves what this or there refers to. Multimodal systems need it to link words to the world.

A transactional query aims to perform an action, such as buying, downloading or booking. It is one of Broder's three classes. Treating it like a request for information sends the person away empty-handed.

A navigational query aims to reach a particular website or destination. The person knows where they want to go. The fastest correct answer is the destination itself.

Pre-intentional communication consists of behaviours that others interpret as meaningful before the person uses them deliberately, the first stage in many developmental models. Intent grows out of being understood. Communication starts with the listener's reading.

Semantic role labelling identifies who did what to whom, when and where in a sentence. It extracts the structure of an event. It captures what a request is about, beyond its emotional tone.

Session-based search treats a sequence of queries as one search with an evolving goal. Intent changes and becomes clearer across the session. A classifier working query by query sees only fragments.

Affective computing, a field named by Rosalind Picard in 1997, develops systems that recognise and respond to human emotion. It treats emotional state as input for adapting the interaction. Intent and emotion travel together.

Aphasia is a language impairment acquired through brain injury, most often a stroke. People with aphasia usually know what they want to say and struggle to say it. Their intent shows that meaning and words can come apart.

Also discussed at https://www.asha.org/public/speech/disorders/aphasia/

Eye gaze tracking measures where a person looks with sensors and cameras. It can reveal interest and intent without any input. Interfaces can use it to anticipate what someone wants.

Described in a Google patent: https://patents.google.com/patent/US8510166B2/en

ISO/IEC 24751 is an international standard for describing individual needs and preferences so that digital resources can adapt to them, particularly for accessibility. It lets a system meet a user's way of communicating. Intent is easier to serve when needs are declared.

A board certified behavior analyst designs and supervises behavioural interventions, including programmes that teach functional communication. The role works alongside speech-language pathology. It connects understanding a behaviour with changing it.

Also discussed at https://www.bacb.com/bcba/

Functional communication training teaches a person to replace a challenging behaviour with a communicative act that achieves the same goal. It starts from the intent behind the behaviour. The goal stays, the signal changes.

A query is one signal of an intent that also shows in context, sequence, tone and what the person does next. Systems that classify single queries into a few categories capture a fraction of it. The rest decides whether the person is served.

Serve the goal behind the query: recognise the kind of request, extract the details it needs, and offer a way forward when the need fits no category. Where the goal is unclear, ask. That is the idea of Supervised Search: the page creates the shared attention that a search box lacks.