Roberto La Rocca, reporter at the Parliamentary Reporting Office of the National Parliament of the Netherlands
Central to this first presentation in Sunday afternoon’s IPRS session is the notion that conceptions of quality of parliamentary reporting that are merely based on use and users’ desires and expectations, as suggested by IPRS in its call for papers for this 2017 IPRS conference, fall short in at least one crucial way. In fact, users of parliamentary records range from politicians, bureaucracy and the press to various kinds of citizens, researchers and others. All of these audiences (user groups/clients/customers/target groups/stakeholders) make use of parliamentary records in various ways, for different reasons, and with diverging and sometimes conflicting purposes and interests. It is simply impossible to please everyone all of the time.
Fortunately, this peripheral parliamentary record user base does not constitute an extraordinarily critical community; either out of ignorance, indifference or confidence, the quality of parliamentary reporting is typically taken for granted. Ironically, professional parliamentary reporters themselves may be considered the most avid, alert, serious and critical ‘users’ of their own products, due to various internal quality control activities such as peer reviews and monitoring. In this vein, the principles and standards applied by parliamentary reporting offices themselves, e.g. those regarding editing, function as de facto requirements, needs, desires and expectations regarding record quality – notwithstanding their official role as providers of a product to a consumer audience.
Parliamentary reporting in practice often involves (heavy) editing of the words actually spoken in parliament. Several societal, political, cultural, technological and media developments combined foster an observable trend towards minimizing editing activities in order to produce an increasingly word-for-word type of rendering of parliamentary proceedings. Still, even in highly mediatized democracies and political cultures that increasingly focus on personalization, political emotion and drama, there is still a need for factual, accessible, reliable, and, crucially, authorized and authenticated accounts of parliamentary deliberation – indeed, professional parliamentary reporters’ traditional core business product. Since innovations such as integral live streaming of debates, social media and automatic speech recognition provide alternative accounts of parliamentary proceedings that are better suited for new purposes, they may even increase the relevance of and demand for a more detached version of parliamentary reports.
Fyke Goorden, Ingeborg Mulders and Susanne Parren
Since five years, the Dutch Parliamentary Reporting Office uses a method of peer evaluation in which all reporters annually check each other’s work against the audio. In this process, the summer recesses are used to have a large amount of turns of all 40 reporters checked by their colleagues.
This is a peer review evaluation process because the feedback they give to each other is based on practical knowledge and experience. Since all reporters are reporting on a daily basis, they know what they are talking about when commenting on each others work. Because everyone is involved in this system of checking turns against the audio, it is a truly horizontal and bottom-up system of monitoring quality. And because of the opportunity to comment on the turns anonymously, this system offers comfort to the reporters and creates a safe space for constructive criticism and feedback. It is meant to appeal to everyone’s intrinsic motivation to deliver a good Report. The goal is to have the reporters draw their own conclusions to improve their work.
Dario Savalli and Fabrizio Verruso
Besides the typical function of parliaments (the legislative one), one cannot ignore another typical and important function: controlling the work of the Executive. This controlling function has become more important to Italy’s and Sicily’s parliaments, because of the enhanced legislative role of the government. The government’s actions, especially those in the administrative field, have to be checked for correctness and regularity. For this, questions and queries are used. In their presentation, Savalli and Varusso briefly explained the steps involved in submitting a question or query. After that, they elaborated on the qualitative inspection activities they and their colleagues carry out along three stages: submission, inspection and follow-up. In these stages, questions and queries submitted need to be checked for the right category, the number of MPs signing, the date, the addresses, correct references to laws etcetera. Only if this information is entered correctly, a fast search into the database, that is accessible for everyone, is possible (www.ars.sicilia.it).
Minori Arai, assistant managing editor at the Canadian Senate, began her presentation with a description of the recruiting process of reporters for the Canadian Senate. She stressed the importance of experience in captioning skills, possessing a broad dictionary and knowledge about the political process. To ensure that candidates meet the high level of professional and behavioural competencies requested by the Canadian Senate, the recruiting process includes a competition, an interview and a real-time exam.
The Canadian Senate uses the information management system IRIS to add information such as legislative documents and events to their reports. Being an XML (eXtensible Markup Language)-format, IRIS is a customizable system. The XML-editor can thus easily index and tag content, providing for instance links to biographies of senators or videos. This enable a feature-rich search based on keywords. It also makes it easy to link content to social media platforms.
Ms Arai pointed out that the development of this system at the Canadian Senate followed the implementation of a similar system at the Canadian House of Commons. Therefore, the development team knew about certain challenges that would arise and was able to prepare for these. This resulted in a successful progressive launch of IRIS. The combination of real-time reporting and IRIS enables Canadian reporters to meet the requests of their clients. It brings more transparency and increases the accessibility of parliamentary data.
Wouter Zwijnenburg, reporter at the Dutch Reporting Office, presents The Language Room, a subdivision of the Dutch Reporting Office, and its efforts to secure some form of unity in the use of language in the reports of the office and all the parliamentary documents.
With their own interactive site, available for everyone working in the parliament, a constantly growing list of words, containing names of people and organizations, and the frequently updated guidelines The Language Room, together with the Quality Division, another subdivision of the office, works on high quality reporting and writing.
The way people speak in daily life and in the media of course influences parliamentary linguistic usage and vice versa. The Language Room tries to guard the balance. Not every newfangled word or sentence has to immediately find its way to the parliamentary reports. On the other hand: outdated language rules have lost their meening. The use of the verbatim report as subtitles at the Dutch Missed Debate site also has its challenges. Now more than ever everyone who watches debates can compare the spoken word to the recorded texts.
In its seven year existence The Language Room has proved to be a reliable and useful source of linguistic rules and regulations. Its authority in the Dutch parliament is undisputed.
Italian scholar of linguistics Carlo Eugeni presented his audience with a guided tour into the world of speech recognition and its infinite possibillities. A common misconception is that speech recognition and voice recognition are the same thing. The difference is essential however, as Carlo made clear. The difference being that anyone can use the standard installed speech recognition software (like Siri) on any mobile phone. Voice recognition software however uses voice biometric verification technology so that only the authorized user has access.
Speech recognition or speech-to-text means translation of the spoken word into digital data. There are solutions for automatic transcription and also for respeaking, like in the well-known Dragon-tool. With this tool the respeaker orally produces another spoken text wich is transcribed into written text. Automatic transcription is also used in live reporting, including the new feature of edited automatic transcription. This gives the possibility to add live corrections. Next to live reporting, there is intralingual reporting. Translations into any chosen language are made live, either human made, semi-automatic or automatic.
Paolo Paravento, the CEO of PerVoice, came to Berlin to present the latest innovation of his company. PerVoice, an Italian-based enterprise, has a trackrecord of over more than 20 years of research and development in speech recognition. PerVoice develops it’s own technology, including over 60 language models, and is mainly servicing other corporations in the business of processing, analysing and evaluating spoken language. It’s main competitor is Nuance, the developer of respeaking tool Dragon.
“Technology is only technology”, Paolo stated, indicating that the way technology is used is of the utmost importance. It must be supportive to workers needs.The latest innovation of PerVoice is the PerVoice Stenotyping Workstation (PSW), a sophisticated tool for realtime Automatic Speech Recognition. PSW can be used in three modes: direct mode, respeaking mode and mixed mode. The respeaking mode can be used in the case of overlapping speakers, a common problem in the field of speech recognition.
Henk-Jan Eras and Deru Schelhaas, both working with the Reporting Office of the Dutch Parliament, went knee deep in the muddy waters of automatic speech recognition (ASR). Their presentation focused on the (perceived) benefits of automatic speech recogniton for parliamentary reporting. Overpromised and underdelivered, this statement summes up their initial thoughts and feelings on ASR. For more than 20 years the predication is that ASR will put reporters out of their job, but the parliamentary reporter as a vocation is still very much alive and kicking.
In a recent experiment two ASR-suppliers, a start-up and a university spin-off, got tasked by the Reporting Office with transcribing carefully chosen audiofragments of some plenary debates. The outcome in speech-to-tekst scores ranged from 85 to 93 percent.
The preliminary findings of this experiment are: 1. Speech-to-text is not up to a good standard for daily use in the Reporting Office, 2. To much editing effort is required and 3. somewhat surprising: even 100% accuracy is not enough! For use in a Parliamentary Reporting Office, a perfect ASR-solution, delivering a 100% accurate audio-transcript, with no editing, is not doable. Therefore the search is not for a speech-to-text but for a speech-to-report solution, including some automated editing. To try and reach this next level the Reporting Office recently started a collaboration with the Radboud University Nijmegen.
The ensuing audience discussion about the Pros and cons of automatic voice and speech recognition (ASR) in reporting generated some accounts of practical experiences with the actual application of ASR in parliamentary reporting contexts. In the Scottish parliament ASR is used as a replacement for typing, with increased production speed and a reduced need for hired external support as positive results. The Italian Senate already offers verbatim reports integrated with video footage of debates. Experiments conducted in Canada with ASR and respeaking turned out to favor the latter, which may be explained by the country’s bilingualism and its sizable population of non-native English speakers. The US-delegation pointed at the opportunities digitization offers for augmented reality and added metadata, e.g. for the benefit of the deaf and visually impaired.
On a more speculative note, some concerns were expressed about the impact of massive digitization and visualization of parliamentary proceedings on democratic politics and society as a whole, and about the consequences of the application of ASR for professional parliamentary reporters. For citizens, increased speed and improved availability of political information may promote empowerment and increase transparency, but superficiality and systemic overload may easily prove to be the unattractive other side of the coin. For reporting professionals, ASR may lead to more interesting, more productive or in other ways more rewarding work, but ASR could equally turn out to be devastating to their job security, professional prospects and job satisfaction.
Whether parliamentary reporters as of 2017, by applying ASR, were making themselves and their profession progressively irrelevant and obsolete, remains to be seen. Holograms at the virtual 75th IPRS-Conference on Mars in 2065 are expected to provide more conclusive data on this issue.
Mr. Treschwig compares the required handwriting quality in longhand and stenography in order to achieve a comparable degree of readability. For this he uses the Latin cursive handwriting system versus the Deutsche Einheitskurzschrift (DEK), the German unified stenography.
Two parameters determine a minimum realization degree of the precision in a writing system:
The Latin cursive handwriting system comprises 59 elements (letters). DEK has 157 elements. So based on the number of elements the realization of the norm is 2.66 times more important in basic DEK than in longhand.
Counting strokes to write ten sentences that are non-interrelated and selected by pure chance from different text sources leads to the following conclusion: Latin cursive handwriting needs 3,166 and DEK needs 918 strokes. So based on the the extend of graphic substance, realization of the norm is 3,166 : 918 = 3.45 times more important in DEK than in Latin cursive handwriting. As both parameters work together one must add both factors: 2.66 + 3.45 = 6.11.
Assuming comparable readability, writing in DEK needs a discipline in precision that is greater than that of Latin cursive writing by a factor 6.
In his presentation, Mr. Kaneko traced back the origins of stenography in Japanese, Chinese and Korean through the history of Western shorthand writing. Abbreviated note writing was established in ancient Rome. It was rediscovered in the Middle Ages, but the real renaissance of stenography began in England in the 18th century. There a basic and systematic stenographic alphabet was designed with the help of geometric theory and techniques for cursive writing. Many nations adopted stenography for publishing verbatim records of parliamentary debates and court hearings, which helped with spreading its use.
The usefulness of abbreviated writing was soon acknowledged in Asia. Western stenographic systems, for instance Gabelsberger and Stolze, were adapted for use in Japan, China and Korea. In the 20th century, the use of electronic stenotype progressed rapidly in Korea and China and Japan followed, with the help of techniques developed in the West.
In the last part of his presentation, Mr. Kaneko shared his experiences with a project about Joseph Schumpeter. Schumpeter used a mix of stenography and longhand writing in the notes for his famous book Capitalism, Socialism and Democracy. Mr. Kaneko meticulously transcribed these notes, resulting in the publication by Joseph Alois Schumpeter, “Supplemental Passage References for Capitalism, Socialism and Democracy”.
Stefan Loeffler, Germany, shared his perspective on the future of graphic shorthand and asked for an open debate. Based on a brief history of shorthand and on his experiences as a (young-generation) teacher, he concluded that the “natural” target group changed and note taking for knowledge workers should be at the center of efforts today. Stefan Loeffler outlined further options for personal use in the modern IT-era. He highlighted the need to reshape didactics to accommodate the new type of learning. He closed with promotion aspects, making clear that shorthand can only survive if organizations come to a common strategy for improving shorthand’s poor public image.
Mr. Hirano presents EPSEMS, a cursive stenography system to write English and Japanese. In the composition of EPSEMS phonetism is the main principle. That distinguishes it from almost all other Japanese stenography systems, in which the syllabic basis alphabet is adopted.
In EPSEMS vowels are expressed by line letters, and consonants are phonetically related in the way of writing unvoiced and voiced sound. EPSEMS contains double and triple consonant letters, which is very effective in writing both English and Japanese. English is easy to read when consonants are expressed firmly. In the major English stenography methods most vowels are omitted. Japanese is a relatively easy to read when vowels are firmly expressed. EPSEMS is designed to be able to write English vowels naturally and rationally. They are written in straight lines and logically categorized into an a-, an i- and a u-group for each related sound. Within one group, the same shape of letters is used, and letters are distinguished only by the difference in length. The distinction between voiced and unvoiced sound in consonants is expressed by the difference in the end of the stroke. Simplicity, clarity and regularity are the main characteristics of EPSEMS. The system, developed for personal use, is easy to learn, easy to write and easy to translate.
Jorge Bravo and Diana Campi from Argentina informed us that the use of shorthand was not limited to reporting activities. In the extensive shorthand writing collection of Miguel Palant, Argentinian parliamentary stenographer and author of multiple books on shorthand, a book titled “Shorthand for Music” by Jean Kutahialian triggered them to delve into the history of the usage of shorthand to record music and dance.
Jorge Bravo pointed out that in the late 19th and early 20th century, authors in several European and American countries searched for a system to replace the classical music notational system. Many thought that the use of shorthand could be helpful to transcribe compositions simultaneously. This resulted in several proposals for musical notation systems based on shorthand, but none of these would enjoy the same wide use as the classical notation system.
Diana Campi gave an overview of the history of attempts to develop a notational system for dance. The main reason for this has always been to make reproduction of a dance production possible. The advent of copyright laws was an extra reason to transcribe dance, as some countries only allowed copyright for a dance piece when it could be registered in written form.
Jorge Bravo emphasized that their presentation was only a very brief synthesis of their research on this topic. In a concluding remark, he told that while the use of shorthand for music and dance has a long history, the various proposed systems based on shorthand are hardly used in contemporary music and dance practices.
Arjan van Hessen
CLARIAH, Telecats/University Twente, Utrecht University
Arjan describes himself as an “experienced evangelist of speech and language technology with a strong focus on Human-Machine-Interaction”. He started his talk with the following question: How can Human Language Technology (HLT) help by producing and improving the Official Report?
For producing speech to text for the Official Report of the House of Representatives in The Netherlands automatic speech recognition (ASR) is not used (yet?). Because up to now it’s not considered fast, cheap or good enough. But for some years the technique of ASR is already used for subtitling the plenary sessions. The human made transcription, the text of the Official Report, is realigned with the audio. And ASR already offers very interesting possibilities for research, retrieval and searching.
The future use of Human Language Technology and ASR might help reporters to speed up their work and to concentrate on the less boring parts. To reach that aim Arjan van Hessen and his researchers are developing an algorithm that learns to transcode ASR output into text that is close to the written version of the spoken utterances. For this Arjan is building a Parliamentary Language Model.
What does Arjan hope to achieve? 100% perfection in ASR won’t be easily reached, but it will help to speed up the work of reporters and to do their work better, for example if they have an editor for correcting the generated text. Furthermore, in the future this developments might automatically generate subtitles based on ASR. They will offer other possibilities for text presentation. It might also increase the search performance and the online access to everything spoken in Parliament, the searched spoken fragments (both audiovisual and text), automatically generated summaries of each item and debate graphics.
Wim Gerbecks and Sander Pasveer
The Velotype keyboard was first used in 1938. The first digital version was introduced in 1983. By that time Wim Gerbecks learned Velotyping at a young age. By pressing several keys simultaneously, complete syllables and words can be made, instead of typing character by character. At the same time, it is a very ergonomical way of typing. For example the thumb is used for many more strokes than just the space bar. Also the mouse of the hand has a function. That means: typing 2-3 times quicker than a qwerty keyboard and with less energy.
The Velotype keyboard is used as a support for deaf and hard of hearing people, for real time subtitling (even remote from home), note taking at company meetings, audio transcription (judicial) and translations. It’s suitable for all platforms. usable in all applications and suitable for more than 30 languages.
Sander Pasveer invented Text on Top. Text on Top is a wireless software and hardware solution that provides real-time captioning on top of any application running on the secondary computer and without interfering with the use of this secondary computer. That offers the possibility to present text on the same screen, so no second screen is needed anymore. Sander invented it especially for deaf and hard of hearing, but it can improve communication for everyone. It functions wireless with a sort of radio technique. It’s compatible with all sorts of keyboards and stenomachines and on al platforms. Also a translation system is integrated in Text on Top.
During the Intersteno Conference Wim and Sander demonstrated the Velotype, Text on Top and an automatic translation via tablets in an excellent manner!
In The Michela stenotype system as a useful support to disabled people Fabio Angeloni and Paolo A. Michela-Zucco from the Italian Senate present an overview of studies and experiments regarding the use and potentials of the Michela keyboard as a writing, speaking or reading device for disabled people.
The world of stenography and that of the disabled have always shown strong interconnections. From the beginning, inventors such as Antonio Michela have been interested in how the methods, services and devices they developed for stenographic purposes could be applied for the empowerment of the disabled, by helping them to overcome the communicative, cultural and social barriers their handicaps entail.
The Michela keyboard, first designed in 1862, is a member of the phonetic keyboard family. With its piano-like appearance, limited number of optimally proportioned keys and ingenious layout, the device has outstanding ergonomic features. The Michela system is capable of representing an impressive standard number of 37 phonemes (26 consonant and 11 vocal sounds), but it is sufficiently versatile to enable expansions that can meet the phonetic needs of almost any natural language. After a customized training period of about 20 to 30 hours, an average person should be able to reach a processing speed of 300 syllables per minute. With some advanced training in specialized methods, including abbreviation techniques, even 400 syllables per minute are feasible, which is sufficient for debate reporting.
Several studies and experiments have shown the Michela keyboard’s suitability and/or potential as a communication device for mute people and people with voice disorders, by adding software capable of vocalizing individual phonemes. By introducing braille features and with the development of an additional orthographic system, it can offer a writing device for the visually impaired and the blind. Furthermore, its potentials as a therapy tool and support for persons with language handicaps, including dyslexia, have been examined. Future developments of the Michela keyboard (may) produce an improved update of the conventional braille terminal or keyboard, its use as a simple MIDI-keyboard and a wearable one-hand-version of the device.
John Vice, editor of debates at the British House of Lords, proposed the question what to report if members of parliament use a prop in their speech. By showing a number of fragments from various parliaments, he invited us to ask ourselves what we would do in these cases. He identified five possible approaches, and the risks that come with every approach.
The first approach is just reporting the words, ignoring the non-verbal act that accompanied these words. While this is a clear and consistent approach, it is likely to be confusing for the reader. Too much context might get lost, which makes it difficult to really understand what has happened by only reading the words uttered. The second approach, which is the one Mr Vice uses himself at the House of Lords, is to make small changes in the words. This tweaking of words is helpful for the reader, but it is a challenge to do this in a totally objective manner. The third approach is to add a description of the event between brackets. This is what is usually done in continental European parliaments. The risk of this approach is similar to that of tweaking words: a description of the event has to be exactly right and complete as well as unbiased. The fourth approach is to only allude to the event without actually describing it. This could still leave the reader with too little contextual information to understand what has happened. The fifth approach is to report everything, as far as possible. Such completeness is difficult to achieve, which brings the risk that a reporter mixes and matches approaches. This does not contribute to the consistency of the report, but at least it is an attempt to give the reader a complete picture of what happened.
Mr Vice concluded with some general remarks on the use of props in parliament. He pointed out that it differs between parliaments and that this is caused by both parliamentary rules about the use of props and the differences between political cultures. Whether the use of props is effective, is impossible to determine in a neutral way. It all depends on the political point of view whether one thinks the use of a prop is rhetorically successful or a rather pathetic gesture.
Eero Voutilainen, from the Reporting Office of the Finnish Parliament, sheds some light on how to develop the linguistic quality of verbatim reports. Linguistic quality, understood as a subjective term, is always subject to some form of regulation: top-down (as in strict rules) or bottom-up (doing what feels right). Essential in the work of a parliamentary reporter is the challenge to maintain an equibrilium. To transform speech into text “something must be changed to keep things as they are”, stated Eero. A reporter can thus fail in two ways: changing to much (failing up) or changing little (failing down). Moreover, a parliamentary reporter has to take different factors into account, for instance the expected needs of target audiences, personal preferences and ideals, parliamentary reporting culture, Reporting Office guidelines, the genre of parliamentary reports and the values and aimes of the Reporting Office.
In the Finnish Record Office the policy is to make as few and subtle alterations as possible. Wrong facts or false citations for instance are not changed, where minor blunders are corrected. In the Finnish case a wide variety of tools has been developed to uphold linguistic quality: an editorial manual, a parliamentary term bank, editorial meetings, inservice training and a language team. Quality development is in progress via regular and systematic feedback.
Mr. Owain Wilkins is reporter in the House of Commons. Welsh, his first language, is one of the oldest living languages in Europe, with 562.000 speakers. Welsh was spoken in the House of Commons for the first time in 1974, when a Welsh MP swore his Oath of Allegiance in his mother tongue. Later, Welsh was granted an official status in the parliamentary rules. Parliamentarians are since allowed to speak in Welsh, next to English.
Nowadays, the language mostly is spoken in the Welsh Grand Committee, which has between three to six meetings per year. The committee is made up of all 40 Welsh MPs and up to 5 other MP’s. Currently, 11 MPs have Welsh as their first language, 5 are Welsh-language learners.
Welsh is a highly phonetic language, which makes firsthand knowledge an absolute must for a reliable report. One colleague of Wilkins in the House of Commons also has proficient knowledge of Welsh for this task. In the coming years the reporting office strives for a more systemic approach towards Welsh in its reporting services, among which is simultaneous translation/interpretation, Welsh-language guidelines and sufficient staff capacity. No wonder, since according to Wilkins the usage of Welsh “will inevatibly grow” due to recent political developments.
A qualitive analysis of post editing in the Japanese and the European Parliament. That is the subject of the presentation of Tatsuya Kawahara, consultant to the Diet. For the Japanese Parliament two goals are identified regarding a report: accuracy and readability. Research in 2007 made clear that in the Diet the word-for-word difference between an audiotranscript and a report averaged on 13%, wich for 93% consisted of simple editing. The guidelines in Japan are clear: fillers and repeats are corrected and correction of grammatical errors is mandatory.
The same research in 2007 revealed striking differences between the reporting practices of the European Parliament and the Japanese Parliament. In Brussels reporters were editing much more than in Tokyo: 20.5% versus 5.9%. Reordening of words and correcting grammatical errors was also much more the case in Europe (19,6% and 20,1%) then in Japan (5,9% and and 7,5%). A more recent comparison has yet to be made, but the data over the last 10 years show that edits in the Japanese Parliament have been reduces with a massive 40%. The reasons for this reduction in edits can be found in internet broadcasting, changing guidelines and the introduction of automatic speech recognition (ASR).
ASR in the Diet has reached an accuracy of around 93%. Of the reporters 80% is said to be happy with the employment of ASR as it means less time and labor spent on reports.
1887 - 2017 All Rights Reserved. Intersteno - International Federation for Information and Communication Processing