Glossary of voice dictation and transcription for lawyers

Assistant mode

Assistant mode is the ablalo feature that rewrites, summarises and changes the tone of dictated text. To do so, only the text travels, encrypted, to a European AI, and that content is not used to train any model. The voice is still transcribed on the computer itself, even when the Assistant is used.

In ablalo: ablalo's privacy policy (in Spanish).

Automatic punctuation

Automatic punctuation is the feature of a dictation program that adds commas, full stops, question marks and capital letters according to the rhythm and meaning of the sentence, without you having to say each mark aloud. In a short message it usually does the job even in free dictation tools; a long document is where the difference shows most. ablalo punctuates in Spanish without you saying «coma» (comma) or «punto y aparte» (full stop, new paragraph).

On the blog: how to dictate in Word.

Cardioid microphone

A cardioid microphone is a directional microphone that mainly picks up the sound that reaches it from the front and much less from the sides and from behind; its pickup pattern, seen from above, is heart-shaped. Pointed at your mouth, it keeps your voice above the noise in the room, unlike the omnidirectional microphones of many laptops.

On the blog: the best microphone for voice dictation.

Cloud versus local processing

Cloud processing is processing done on a provider's servers, to which the audio or the text is sent; local processing is done on the computer itself, without that data leaving the device. For a law firm, the difference determines who ends up with a copy of the client's information and in which country it is processed.

On the blog: transcribing audio to text in a law firm and what to ask before using AI with client data.

Consolidated text («texto consolidado»)

A consolidated text is the version of a law that brings together in a single text the amendments, corrections and express repeals it has had since it was first enacted. The BOE (the Spanish Official State Gazette) offers it to make the law in force easier to access, and warns that it is for information only and has no legal value. ablalo's legal database inserts articles from the BOE consolidated text.

Source: BOE, what a consolidated text is (in Spanish).

Custom dictionary

A custom dictionary, or personal vocabulary, is the list of words a user adds to a dictation program so that it recognises them and spells them correctly: uncommon surnames, company names, acronyms or the firm's jargon. Windows and Word dictation do not support one. In ablalo you add them once and, if you correct the same word twice, it offers to learn it.

On the blog: ablalo vs free Windows and Word dictation.

Data controller («responsable del tratamiento»)

A data controller is, under Article 4(7) of the GDPR, “the natural or legal person, public authority, agency or other body which, alone or jointly with others, determines the purposes and means of the processing of personal data”. It is whoever decides why and how personal data is processed. For the data a law firm processes with ablalo, the controller is the law firm.

On the blog: professional secrecy and the GDPR when using AI in a law firm. In ablalo: who is responsible for each piece of data in ablalo (in Spanish). Source: BOE, Regulation (EU) 2016/679 (GDPR), Article 4(7) (in Spanish).

Data processor («encargado del tratamiento»)

A data processor is, under Article 4(8) of the General Data Protection Regulation (GDPR), “a natural or legal person, public authority, agency or other body which processes personal data on behalf of the controller”. Its relationship with the controller is governed by Article 28 of the GDPR. ablalo acts as the processor of the data the law firm processes with the app.

On the blog: professional secrecy and the GDPR when using AI in a law firm. In ablalo: ablalo's privacy policy (in Spanish). Sources: BOE, Regulation (EU) 2016/679 (GDPR), Articles 4(8) and 28; AEPD (the Spanish Data Protection Agency), guidelines for contracts between controllers and processors (both in Spanish).

Diarisation

Diarisation, or speaker diarisation, is the technique that splits a recording according to who is speaking at each moment: it detects changes of voice and groups together the segments from the same person. It does not identify anyone by name; it labels each one as “Speaker 1”, “Speaker 2” until someone gives them names. In ablalo, the Hearings tab separates the speakers in a hearing or a meeting and lets you name them.

In ablalo: meeting and hearing transcription with ablalo.

Dictaphone

A dictaphone is a portable voice recorder designed for dictation: the professional records what they want to write and another person, or a program, later turns it into text. It is the classic workflow of many law firms, with transcription queues and support staff. With live dictation, the text appears as you speak and that intermediate step disappears.

On the blog: ablalo vs Dragon Legal.

European AI

A European AI is an artificial intelligence service provided by a company established in the European Union and subject to the GDPR. Where the data is processed, and whether its providers transfer it to third countries such as the United States (Chapter V of the GDPR), is checked in its data processing agreement and its privacy policy. In ablalo's Assistant mode, the text travels encrypted to a European AI.

In ablalo: ablalo's privacy policy (in Spanish). Source: BOE, Regulation (EU) 2016/679 (GDPR), Chapter V (in Spanish).

Hearing recording («grabación de vistas»)

Hearing recording is the video and audio recording of oral proceedings before a court. Article 147 of the Ley de Enjuiciamiento Civil (the Spanish Civil Procedure Act) provides that oral proceedings in hearings and appearances («vistas, audiencias y comparecencias») held before judges or magistrates, or before the court clerks («letrados de la Administración de Justicia»), are recorded on a medium capable of recording and reproducing sound and image.

In ablalo: transcribe a hearing on your own computer. Source: BOE, Ley de Enjuiciamiento Civil, Article 147 (in Spanish).

Libro de estilo de la Justicia

The Libro de estilo de la Justicia (the Spanish justice style guide) is the work of the Real Academia Española (the Royal Spanish Academy) and the Consejo General del Poder Judicial (the General Council of the Judiciary), published in 2017, that sets out writing criteria for legal texts: paragraph organisation, punctuation, Latin expressions, capital letters and numerical expressions, with a vocabulary of correct usage. In the legal register, ablalo applies its criteria, for example «con base en» instead of «en base a» (both meaning “based on”).

On the blog: common mistakes in legal writing according to the Libro de estilo. Source: RAE, Libro de estilo de la Justicia (in Spanish).

Local transcription

Local, or on-device, transcription is transcription done by the computer itself with a speech model installed on it, without sending the audio to any server. It usually works offline and the recording does not leave the device. Mac dictation uses it on computers with Apple silicon for supported languages; in ablalo it is how it works by default, not an optional mode.

On the blog: how to turn on dictation on a Mac and transcribing audio to text in a law firm.

Minutes («acta»)

The minutes of a meeting or a proceeding are the document that records what happened: who took part, what was discussed and what was agreed. For a meeting they usually record the essentials, not every word, so they are not the same as a transcript. In the courts, Article 146 of the Ley de Enjuiciamiento Civil (the Spanish Civil Procedure Act) provides that procedural acts that do not consist of written submissions and documents are documented by means of minutes and records («actas y diligencias»).

Source: BOE, Ley de Enjuiciamiento Civil, Article 146 (in Spanish).

Noise cancellation

Noise cancellation is the set of techniques that reduce the background sound a microphone picks up (the air conditioning, the keyboard, a nearby conversation) so that mostly the voice is heard. The microphone itself or a program on the computer can do it. For dictation, very aggressive cancellation, designed for video calls, can cut out nuances of the voice that the recognition needs.

On the blog: which microphone to choose for voice dictation.

Professional secrecy («secreto profesional»)

A lawyer's professional secrecy is the duty to keep confidential what they learn through their work. Article 542.3 of the Ley Orgánica del Poder Judicial (the Spanish Organic Law of the Judiciary) puts it as follows: “Lawyers must keep secret all the facts or information they learn by reason of any of the forms of their professional activity, and may not be compelled to testify about them” (our translation).

On the blog: professional secrecy and the GDPR when using AI in a law firm. Source: BOE, Ley Orgánica del Poder Judicial, Article 542 (in Spanish).

Speech model

A speech model is the artificial intelligence program, trained on large amounts of audio and its text, that turns speech into written text. It can run on the provider's servers, in which case the audio leaves the device, or on the computer itself, if the model is downloaded and works offline. In ablalo, the speech model runs on the user's computer.

In ablalo: ablalo's privacy policy (in Spanish).

Speech recognition

Speech recognition, or automatic speech recognition, is the technology that turns the voice into written text: it analyses the sound, decides which words are the most likely and writes them. It is the basis of dictation, of the transcription of recordings and of voice assistants. It is not the same as speaker recognition, which is used to identify who is speaking.

SRT and VTT

SRT and VTT are two plain-text subtitle file formats: each block states the moment it starts and ends, in minutes, seconds and milliseconds, and the text spoken in that interval. SRT comes from the SubRip program; VTT, or WebVTT, is the format the W3C defines for video subtitles on the web. Both open in any text editor.

Source: W3C, WebVTT: The Web Video Text Tracks Format.

Textualización

Textualización is the Spanish Ministry of Justice system, called Textualización de grabaciones (textualisation of recordings), that automatically extracts the text of what is said in recorded trials, hearings and appearances. It is based on neural learning techniques and is integrated with the recording systems of the courtrooms. It lets you search by keyword, see the speakers identified by a number and download the text as an editable document.

Source: Ministerio de Justicia (the Spanish Ministry of Justice), Textualización de grabaciones (in Spanish).

Timestamps

Timestamps are the indication of the minute and second at which each part of a recording is said. In a transcript they let you go back to the exact moment in the audio or video to check or quote a sentence. ablalo syncs the text with the video, saves notes with the exact second and exports the transcript to Word with timestamps.

In ablalo: ablalo's Hearings tab.

Transcription

Transcription is turning an existing audio or video recording into text: a hearing, a statement, a meeting or a voice note. It can be verbatim, with every word as it was said, or clean, without hesitations or repetitions. Unlike dictation, the text is not produced while you speak but from the file, and it usually shows who is speaking and at what minute.

On the blog: transcribing audio to text in a law firm. In ablalo: meeting and hearing transcription with ablalo.

Verbatim transcription

Verbatim transcription reproduces every word as it was said, with filler words, repetitions, broken sentences and corrections halfway through a sentence. It is what you need when exactly what was said, and how, matters. For drafting a document it is the opposite of what you are after: when you dictate, ablalo gives you back text that is already written up, without filler words or repetitions, instead of a verbatim transcript.

Voice commands

Voice commands are the words or phrases that the dictation program interprets as an instruction instead of typing them: «nuevo párrafo» (new paragraph), «abrir paréntesis» (open bracket) or «borra eso» (delete that). In classic dictation they were also used to insert each punctuation mark. In ablalo you dictate, for example, «nuevo párrafo», «primero punto guion» (which writes Primero.-) or «abrir mayúsculas» (capitals on).

On the blog: the Windows 11 dictation commands in Spanish.

Voice dictation

Voice dictation is a way of writing by speaking: you speak and the program types the text, live, wherever the cursor is. It differs from transcription in that it is for producing a new text, not for turning an existing recording into text. Current programs punctuate on their own and some, like ablalo, also remove filler words and repetitions.

On the blog: the best dictation software for lawyers in Spain.

Voice training

Voice training is the initial session in which the user reads texts aloud so that the dictation program creates a profile adapted to their voice and pronunciation. It was common in earlier generations of dictation programs. According to Nuance, Dragon no longer needs it, and neither does ablalo: it works from the first minute, without creating a voice profile.

On the blog: ablalo vs Dragon Legal.

Voice typing

Voice typing is Microsoft's name for the free dictation in Windows 10 and 11: with the cursor in a text box you press Windows + H and what you say is typed in any program. It uses online speech recognition, so the audio is processed on Microsoft's servers and it does not work offline. It does not support a custom vocabulary.

On the blog: voice typing in Windows 11. Source: Microsoft, Use voice typing to talk instead of type on your PC.

Word error rate (WER)

Word error rate (WER) is the usual measure of the accuracy of speech recognition: it adds up the words substituted, omitted and inserted compared with a reference transcript made by a person and divides the result by the number of words in that reference. The lower, the better: 5% means about five wrong words in every hundred.

Words per minute

Words per minute is the measure of speed when writing or speaking: how many words are typed, dictated or spoken in a minute. The reference we use is that a person types about 40 when writing and speaks fluently at about 160, so dictating the draft is roughly four times faster. The review afterwards is not included in that sum.

On the blog: how much time you save by dictating instead of typing.

Zero Data Retention

Zero Data Retention is the contractual commitment by which an AI provider does not keep what it receives or what it returns. ablalo has it in its contract with its AI provider; for its part, ablalo keeps those requests for 14 days in the account to review the quality of the assistant, unless you ask it not to. The Enterprise plan includes an enforced privacy mode with zero data retention.

In ablalo: how long each piece of data is kept in ablalo (in Spanish).