Traditional NLG pipeline and NLG for news

We cover a grounding in the traditional stages of the NLG pipeline, through to recent advances, with a focus on news.

This course will begin with an overview of the typical technical architecture of a Natural Language Generation (NLG) system and some of the techniques used at different stages of the pipeline. In the following sessions, each presented by a different participant, we will look more closely at recent advances in different technologies that form parts of the traditional pipeline, as well as more recent alternatives to the pipeline, including NLG using neural networks.

The focus of the course will be on application to automation of parts of the production of journalistic news content. We will consider the application of the techniques considered to these tasks, research specifically concerning this process and existing systems already in use in this domain.

Each session will consist of a 45 min presentation on a different subject, followed by group discussion. The first presentation, on the traditional NLG pipeline, will be given by Mark Granroth-Wilding. Each participant will present at one of the following sessions. Topics and some relevant literature are available below.

Ilmoittaudu

Viestit

Mark Granroth-Wilding

Julkaistu, 6.2.2017 klo 12:10

Dear NLG group participants,
Tomorrow we'll be looking in more depth at the traditional NLG pipeline, with presentations from Myriam, Julian and Aleksi.

Some reading materials are up on the course page for you to have a look over before then (see "Week 4" section).

See you tomorrow,
Mark

Mark Granroth-Wilding

Julkaistu, 30.1.2017 klo 13:05

Hi all,
Tomorrow, we will be meeting for week 3 of the news NLG study group. We will have presentations from Gusse, Lauri, Heikki and Eero.

Some references and an abstract are now available on the course webpage (see the section for week 3). Please take a look over some of the material, so that you're able to contribute to the discussion tomorrow.

Remember that replying to this email won't reach me: email me personally if you want to get in touch.

Looking forward some lively and illuminating discussion tomorrow!
All the best,
Mark

Mark Granroth-Wilding

Julkaistu, 23.1.2017 klo 15:45

To participants of the NLG for news automation study group,

It was good to see so many of you last week at the introductory study group meeting. Tomorrow we get things going with some short introductions to the different topics we will be covering later in the course.

Hopefully, you've all had time now to take a first look over some of the literature in your topic. Tomorrow, you should be prepared to talk for a few minutes on some ideas you have so far about what you will investigate. Obviously, we don't expect you to be able to give any detail yet, just a overview of roughly what's coming.

Think about the following:
- What areas of your topic are you most interested in pursuing? (You'll find it much easier to do the reading if it's something you're interested in!)
- Does any of the suggested reading look particularly interesting?
- What other literature have you found that's especially relevant to our application?

You don't need to prepare slides for these presentations -- they're too short for it to be worthwhile.

Remember that the topic description and allocations, as well as initial suggested reading, are on the course webpage: https://courses.helsinki.fi/582767/117339112.

Contact me if you have any problems. See you all at 10:15 tomorrow.

All the best,
Mark

Aikataulu

Tästä osiosta löydät kurssin opetusaikataulun. Tarkista mahdolliset muut aikataulut kurssiesitteestä.

PäivämääräAikaOpetuspaikka
Ti 17.1.2017
10:15 - 12:00
Ti 24.1.2017
10:15 - 12:00
Ti 31.1.2017
10:15 - 12:00
Ti 7.2.2017
10:15 - 12:00
Ti 14.2.2017
10:15 - 12:00
Ti 21.2.2017
10:15 - 12:00
Ti 28.2.2017
10:15 - 12:00

Materiaalit

Topic descriptions

Topics for each week. For suggested initial reading lists, see the sections below.

Week 1: Introductory talk.
Slides: https://courses.helsinki.fi/sites/default/files/course-material/4476853/...
- Overview of classical NLG pipeline, based on Building NLG Systems

Week 2: Short overviews
- Everyone presents a very short overview of some of the things we will cover in their session
- Not had much time to read yet, but should have had a chance to glance over some literature
- We discuss what we’re most interested in in each topic, so we know what to focus on when preparing

Week 3: Goals and architecture of algorithmic journalism
- Some high-level material about what algorithmic journalism can be (or already is) useful for
- Hopefully, end up with some idea of possible architectures of automated journalism systems
- Template-based news NLG. This is the predominant approach to news automation at the moment
- Review some existing approaches (e.g. Wordsmith) and how they go beyond just simple templates

Week 4: NLG pipeline for weather reporting
- Classical NLG pipeline, going into more detail on some of the components w.r.t. the specific example of weather reporting
- There’s lots of literature on this, as it’s a task that’s received a lot of attention in NLG
- Make sure to cover recent literature on specific sub-tasks – not just example given in R&D
- Limitations of these approaches when going beyond weather (in particular, considering news)

Week 5 & 6: Statistical NLG (split into two)
- Text-to-text generation and how it fits into the pipeline
- Some statistical approaches to individual pipeline stages, e.g. referring expression generation
- Neural NLG
* This is a hot topic. As it’s an active area of research, approaches in the literature will be less well established, so pick some basic examples to work through
* No mathematical details of neural networks: will take too long and distract us
* How might these approaches be used in a news automation context?
* What are the problems with using end-to-end machine-learned approaches to generate news?
* Can we combine NN approaches with traditional pipeline? How?
- Where in pipeline is machine-learning appropriate or potentially most useful? Where might we consider using more traditional, labour-intensive approaches?

Week 7: Interaction with journalists
- If we choose not to build an end-to-end generation system (as in IA project), where do journalists fit into the pipeline? (Or, where can NLG fit into journalism?)
- Some examples of semi-automatic journalism and tools for assisting journalists. Wordsmith is one, but it’s not very interesting and it would be good to cover other sorts of tools (and not overlap with template session)
- What parts of journalistic process are crying out for automation? Could NLG help?
- What parts of the process are not worth automating or too hard?

Topic allocation

During the first session, we decided what topic/sub-topic each participant will investigate and present.

Week 3: Goals and architecture of algorithmic journalism
- Carl-Gustav L: overview
- Lauri H: modelling the structure of articles
- Eero L: sports journalism
- Heikki R: something else

Week 4: NLG pipeline for weather reporting
- Aleksi: content determination
- Julian S: referring expression generation and realisation
- Myriam M: lexicalisation and aggregation

Week 5: Statistical NLG, part I: statistical approaches to pipeline components
- Kari K: aggregation
- Jared: linguistic and structure realization
- Keith: other components in the microplanner

Week 6: Statistical NLG, part II: alternative architectures
- Nikola M: neural networks (split with Keith)
- Keith D: LSTMs and GANs
- Leo L: intro to probabilistic approaches
- Aaro S: probabilistic modelling and end-to-end architectures

Week 7: Interaction with journalists
Division of topic to be discussed.
- Otto B: research tools for journalists
- Johanna J: why are there still so many jobs in journalism?
- Laura: journalists in the pipeline

Week 3 topic: Goals and architecture of algorithmic journalism

Presentations by Gusse, Lauri, Eero and Heikki.

References from Eero:
Carlson, M. (2015). The robotic reporter: Automated journalism and the redefinition of labor, compositional forms, and journalistic authority. Digital Journalism, 3(3), 416-431.
Graefe, A. (2016). Guide to automated journalism.
Dörr, K. N. (2016). Mapping the field of Algorithmic Journalism. Digital Journalism, 4(6), 700-722.
Clerwall, C. (2014). Enter the robot journalist: Users' perceptions of automated content. Journalism Practice, 8(5), 519-531.
Van Dalen, A. (2012). The algorithms behind the headlines: How machine-written news redefines the core skills of human journalists. Journalism Practice, 6(5-6), 648-658.
BBC: Robo-journalism: How a computer describes a sports match http://www.bbc.com/news/technology-34204052
https://techcrunch.com/2016/07/03/ap-sports-is-using-robot-reporters-to-...
Wired: Wordsmith's 'robot journalist' has been unleashed http://www.wired.co.uk/article/wordsmith-robot-journalist-download
http://www.theverge.com/2016/7/4/12092768/ap-robot-journalists-automated...
http://www.poynter.org/2010/statsheet-technology-generates-game-stories-...
http://www.poynter.org/2010/statsheet-network-automates-hundreds-of-spor...

Abstract (including references) from Heikki:
https://courses.helsinki.fi/sites/default/files/course-material/4478814/...

------------------------
Some starting points for reading in this topic. You do not have to use all of these, they are just suggestions. You should certainly also add further literature, for example by looking at references in these articles or searching for relevant papers.

Article: Mapping the field of Algorithmic Journalism, Dörr (2015)
http://www.tandfonline.com/doi/abs/10.1080/21670811.2015.1096748 (available through university)
Review of automatic journalism from a high-level perspective (not focused on technical material).

Report: Guide to Automated Journalism, Andreas Graefe (2016)
http://towcenter.org/research/guide-to-automated-journalism/
Recent review of the field. Very useful, both for its content and reference list as a source of further literature.

Commercial system: Wordsmith (Automated Insights)
https://automatedinsights.com/wordsmith
One of the most commonly used systems for automated journalism today. Largely template-based, old-fashioned NLG.

Popular article: How a robot wrote for Engadget, Souppouris (2016)
https://www.engadget.com/2016/08/15/robot-journalism-wordsmith-writer/
Article about use of Wordsmith in news production.

Article: Summarising News Stories for Children, Macdonald and Siddharthan (2016)
http://www.macs.hw.ac.uk/InteractionLab/INLG2016/proceedings/pdf/INLG01.pdf
Recent study in news generation / summarisation. For this topic, do not focus too much on the technical details: we may look more into those in the statistical NLG topic.

Popular press: Building a Robot Journalist, Espen Waldal
https://medium.com/bakken-b%C3%A6ck/building-a-robot-journalist-171554a6...
Recent popular article (very high-level) about algorithmic journalism.

Report: Guide to Automated Journalism, Andreas Graefe (2016)
http://towcenter.org/research/guide-to-automated-journalism/
Recent review of the field. Very useful, both for its content and reference list as a source of further literature.

Week 4 topic: NLG pipeline for weather reporting

Reading material supplied by the presenters is available here. Below that are the starting points for reading provided by Mark at the beginning of the course.

References from Julian:
Article: Choosing words in computer-generated weather forecasts, Reiter et al. (2005)
Article: Statistical Natural Language Generation from Tabular Non-textual Data, Mahapatra et al. (2016)
Book: Building Natural Language Generation Systems, Dale, Reiter (2000)

References from Myriam:
Goldberg, E., Driedger, N., & Kittredge, R. I. (1994). Using natural-language processing to produce weather forecasts. IEEE Expert, 9(2), 45-53.
Stede, M. (1994). Lexicalization in natural language generation: A survey. Artificial Intelligence Review, 8(4), 309-336.
The following also appear under Julian's, above:
[ Reiter, E., Sripada, S., Hunter, J., Yu, J., & Davy, I. (2005). Choosing words in computer-generated weather forecasts. Artificial Intelligence, 167(1-2), 137-169. ]
[ Book: Reiter, E., & Dale, R., (2000). Building natural language generation systems. Cambridge: Cambridge university press. ]

----------------
Some starting points for reading in this topic. You do not have to use all of these, they are just suggestions. You should certainly also add further literature, for example by looking at references in these articles or searching for relevant papers.

Book: Building Natural Language Generation Systems, Reiter & Dale (2000)
The standard textbook on NLG, now somewhat out of date, but an excellent introduction. It uses an automatic weather report generation system as a running example throughout the book.

Article: A Case Study: NLG meeting Weather Industry Demand for Quality and Quantity of Textual Weather Forecasts, Sripada et al. (2014)
http://www.aclweb.org/anthology/W14-4401
A recent practical NLG system for weather forecasts. Uses commercial NLG software.

Report: WAG: Sentence Generation Manual, O'Donnell (2006)
http://www.wagsoft.com/Papers/Wag.pdf
Manual for a surface realisation system. Unfortunately, the software itself is no longer available (and quite old, anyway), but the manual may give some insight into how the final stages of the traditional pipeline can be implemented in practice.

Marketing: Weather Reporting Case Study, Arria NLG
https://www.arria.com/wp-content/uploads/2015/03/ARR0015F_Weather-Report...
No details of the system at all, but one example of a commercial application of NLG to this task.

Article: Choosing words in computer-generated weather forecasts, Reiter et al. (2005)
http://www.sciencedirect.com/science/article/pii/S0004370205000998
Lexical choice for weather reporting. Mainly interesting for the concrete system description.

Article: Statistical Natural Language Generation from Tabular Non-textual Data, Mahapatra et al. (2016)
http://www.macs.hw.ac.uk/InteractionLab/INLG2016/proceedings/pdf/INLG24.pdf
Recent statistical NLG system focusing on weather reporting. You may wish to leave the more statistical aspects of this for the next session's presenters.

Week 5 topic: Statistical NLG, part I: statistical approaches to pipeline components

Here is reading material submitted by this week's presenters. Below you can find the original suggested starting points for reading.

References from Kari:
A Statistical NLG Framework for Aggregated Planning and Realization. Ravi Kondadadi, Blake Howald and Frank Schilder
Aggregation via Set Partitioning for Natural Language Generation. Regina Barzilay, Mirella Lapata
Building Applied Natural Language Generation Systems (pages 20-22). Ehud Reiter, Robert Dale
SimpleNLG: A realisation engine for practical applications. Albert Gatt and Ehud Reiter

References from Jared:
Emiel Krahmer and Mariët Theune. 2010. Probabilistic Approaches for Modeling Text Structure and Their Application to Text-to-Text Generation
Regina Barzilay and Lillian Lee. 2004. Catching the Drift: Probabilistic Content Models, with Applications to Generation and Summarization
Ehud Reiter and Robert Dale. 2000. Building Natural Language Generation Systems. Cambridge University Press, New York, NY, USA.
Ani Nenkova. 2005. Automatic Text Summarization of Newswire. In Proceedings of the 20th national conference on Artificial intelligence - Volume 3 (AAAI'05), Vol. 3. AAAI Press 1436-1441.

Core references from Keith:
Learning what to say and how to say it: Joint optimisation of spoken dialogue management and natural language generation - Lemon
Learning Lexical Alignment Policies for Generating Referring Expressions in Spoken Dialogue Systems - Janarthanam and Lemon
Context-Based Word Acquisition for Situated Dialogue in a Virtual World - Qu and Chai
Trainable Sentence Planning for Complex Information Presentation in Spoken Dialog Systems - Stent, Prasad, and Walker

Keith's full reference list:
Modelling and Evaluation of Lexical and Syntactic Alignment with a Priming-Based Microplanner - Buschmeier, Bergmann, and Kopp. From Empirical Methods in Natural Language Generation
Natural Language Generation as Planning under Uncertainty for Spoken Dialogue Systems - Rieser and Lemon
Building Natural Language Generation Systems -Chapter 5, Microplanning
Statistical Natural Language Generation from Tabular Non-textual Data - Mahapatra, Naskar, and Bandyopadhyay
Planning Natural Language Referring Expressions - Appelt
Natural Language Generation as Incremental Planning Under Uncertainty: Adaptive Information Presentation for Statistical Dialogue Systems - Rieser, Lemon, and Keizer
Learning Lexical Alignment Policies for Generating Referring Expressions in Spoken Dialogue Systems - Janarthanam and Lemon
Learning what to say and how to say it: Joint optimisation of spoken dialogue management and natural language generation - Lemon
Context-Based Word Acquisition for Situated Dialogue in a Virtual World - Qu and Chai
Collective Content Selection for Concept-To-Text Generation - Barzilay and Lapata
Statistical Acquisition of Content Selection Rules for Natural Language Generation - Duboue and McKeown
Trainable Sentence Planning for Complex Information Presentation in Spoken Dialog Systems - Stent, Prasad, and Walker
Microplanning with communicative intentions- The SPUD system - Stone, Doran, Webber, Bleam, and Palmer
Practical Issues in Automatic Documentation Generation (PLANDoc system)- McKeown, Kukich, and Shaw
SUMTIME-MOUSAM: Configurable Marine Weather Forecast Generator - Sripada, Reiter, and Davy
Automatic Evaluation of Referring Expression Generation Using Corpora - Gupta and Stent
Towards Integrated Microplanning of Language and Iconic Gesture for Multimodal Output - Kopp, Tepper, and Cassell
Unsupervised Concept-to-text Generation with Hypergraphs - Konstas and Lapata
Hybrid Reinforcement/Supervised Learning for Dialogue Policies from COMMUNICATOR data - Henderson, Lemon, and Georgila
A framework for creating natural language descriptions of video streams - Khan, Harbi, and Gotoh
Data-to-text summarisation of patient records- Using computer-generated summaries to access patient histories - Scott, Hallett, and Fettiplace
Automatic generation of natural language nursing shift summaries in neonatal intensive care: BT-Nurse - Hunter et al.
Automatically Learning Cognitive Status for Multi-Document Summarization of Newswire - Nenkova, Siddharthan, McKeown
Learning What to Talk About in Descriptive Games - Zaragoza and Li

--------------------------
Some starting points for reading in this topic.

Article: Summarising News Stories for Children, Macdonald and Siddharthan (2016)
http://www.macs.hw.ac.uk/InteractionLab/INLG2016/proceedings/pdf/INLG01.pdf
Recent study in news generation / summarisation. Perhaps look in more detail here at the technical details and use of summarisation techniques. (Higher level details probably covered already in Algorithmic Journalism topic.)

Book: Empirical methods in natural language generation : data-oriented methods and empirical evaluation, Krahmer & Theune (2010)
https://helka.finna.fi/Record/helka.2380001 (eBook available through university)
Collection of papers describing recent statistical advances in NLG. Many of the papers have a very specific focus, so will not be useful, but the introduction and some of the papers provide a useful update on Reiter & Dale. Look particularly at chapters on text-to-text generation and referring expression generation.

Look at proceedings papers from recent years of the International Natural Language Generation Conference (INLG) and ENLG. Many of these focus on subtasks of the classical NLG pipeline that we've seen, typically applying modern machine learning techniques.

Article: Statistical Natural Language Generation from Tabular Non-textual Data, Mahapatra et al. (2016)
http://www.macs.hw.ac.uk/InteractionLab/INLG2016/proceedings/pdf/INLG24.pdf
Recent statistical NLG system focusing on weather reporting. May have been covered in the previous session, but the statistical approach may be of interest here.

Week 6 topic: Statistical NLG, part II: alternative architectures

Here is reading material submitted by this week's presenters. Below you can find the original suggested starting points for reading.

References from Aaro:
Ioannis Konstas and Mirella Lapata. 2012. Unsupervised concept-to-text
generation with hypergraphs. http://www.aclweb.org/anthology/N12-1093
Belz. 2008. Automatic Generation of Weather Forecast Texts Using Comprehensive Probabilistic Generation-Space Models. (Ties nicely to the week 4)
Kim & Mooney. 2010. Generative alignment and semantic parsing for learning from ambiguous supervision.
(Ioannis Konstas and Mirella Lapata. 2012. Concept-to-text generation via discriminative reranking. http://www.aclweb.org/anthology/P12-1039)
(Ioannis Konstas and Mirella Lapata. 2013. Inducing Document Plans for Concept-to-Text Generation. http://www.aclweb.org/anthology/D13-1157)
(Ioannis Konstas and Mirella Lapata. 2013. A Global Model for Concept-to-Text Generation http://www.jair.org/media/4025/live-4025-7407-jair.pdf)

NB: The four Konstas & Lapata papers are all variations / updates of the same core model of the first 2012 paper.

Abstract from Leo: https://courses.helsinki.fi/sites/default/files/course-material/4481775/...
References from Leo:
P. Liang, M. I. Jordan, and D. Klein, “Learning semantic correspondences with less supervision,” in Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL. 2009
G. Angeli, P. Liang, and D. Klein, “A simple domain-independent probabilistic approach to generation,” in Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2010, pp. 502–512.

-------
Some starting points for reading in this topic. You do not have to use all of these, they are just suggestions. You should certainly also add further literature, for example by looking at references in these articles or searching for relevant papers.

Book: Empirical methods in natural language generation : data-oriented methods and empirical evaluation, Krahmer & Theune (2010)
https://helka.finna.fi/Record/helka.2380001 (eBook available through university)
Collection of papers describing recent statistical advances in NLG. Many of the papers have a very specific focus, so will not be useful, but the introduction and some of the papers provide a useful update on Reiter & Dale.

Look at proceedings papers from recent years of the International Natural Language Generation Conference (INLG) and ENLG. Many of these focus on subtasks of the classical NLG pipeline that we've seen, typically applying modern machine learning techniques.

Article: A Simple Domain-Independent Probabilistic Approach to Generation, Angeli et al. (2010).
http://nlp.cs.berkeley.edu/pubs/Angeli-Liang-Klein_2010_Generation_paper...
A relatively simple end-to-end architecture that models the whole pipeline as a Bayesian generative model. Look at how the pipeline contrasts with those in Reiter & Dale and how it is divided into components.

Article: Unsupervised concept-to-text generation with hypergraphs, Konstas & Lapata (2012)
http://www.aclweb.org/anthology/N12-1093
A good example of the traditional approach of concept-to-text generation, with a pipeline based on machine learning. Their model is logically divided into components corresponding to modules of the traditional pipeline, but training inference and generation treat them as a joint model (i.e. perform them in parallel).

Article: Automatic label generation for news comment clusters, Aker et al. (2016)
http://www.macs.hw.ac.uk/InteractionLab/INLG2016/proceedings/pdf/INLG10.pdf
Really a summarization task, but potentially relevant to news automation, which in some circumstances may make heavy use of summarization. Also good for its focus on news. It doesn't fit into the traditional pipeline in an obvious way.

Article: Context-aware Natural Language Generation with Recurrent Neural Networks, Tang et al. (2016, Arxiv)
https://arxiv.org/abs/1611.09900
An example of neural generation. Quite heavy on the technical details, but one example of the sort of thing that it would be good to cover (at a high level, without maths!).

Article: Stochastic Language Generation in Dialogue using Recurrent Neural Networks with Convolutional Sentence Reranking, Wen et al. (2015)
Another particularly nice example of how to do NLG using neural networks. Note the same again about mathematical details.

Week 7 topic: Interaction with journalists

Here is reading material submitted by this week's presenters. Below you can find the original suggested starting points for reading.

Laura's abstract: https://courses.helsinki.fi/sites/default/files/course-material/4482375/...

References from Laura:
ANDERSON, C.W., 2012. Towards a sociology of computational and algorithmic journalism. New Media & Society, 15(7), pp. 1005.
CARLSON, M., 2015. The Robotic Reporter: Automated journalism and the redefinition if labor, compositional forms, and journalistic authority.l. Digital Journalism, 3(3), pp. 416.
DÖRR, K.N., 2015. Mapping the field of Algorithmic Journalism. Digital Journalism, , pp. 700.
GYNNILD, A., 2014. Journalism innovation leads to innovation journalism: The impact of computational exploration of changing mindsets. Journalism, 15(6), pp. 713.
LECOMPTE, C., 2015. Automation in the Newsroom.
RIORDAN, K., 2014. Accuracy, Independence, and Impartiality: How legacy media and digital natives approach standards in the digital age. Oxford: Reuters Institute.

References from Otto:
Guide to Automated Journalism. http://towcenter.org/research/guide-to-automated-journalism/
The Relevance of Algorithms -Tarleton Gillespie. http://iasc-culture.org/THR/channels/Infernal_Machine/wp-content/uploads...
Clarifying Journalism’s Quantitative Turn - Mark Coddington. http://nca.tandfonline.com/doi/full/10.1080/21670811.2014.976400?src=recsys
The Robotic Reporter - Matt Carlson. http://www.tandfonline.com/doi/abs/10.1080/21670811.2014.976412
A Functional Roadmap for Innovation in Computational Journalism - Nicholas Diakopoulos. http://www.nickdiakopoulos.com/2011/04/22/a-functional-roadmap-for-innov...
Building a Robot Journalist – Espen Waldal. https://medium.com/bakken-b%C3%A6ck/building-a-robot-journalist-171554a6...
Summarising News Stories for Children - Iain Macdonald Advaith Siddharthan. http://www.macs.hw.ac.uk/InteractionLab/INLG2016/proceedings/pdf/INLG01.pdf

--------------
Original starting points for reading in this topic.

Popular press: Building a Robot Journalist, Espen Waldal
https://medium.com/bakken-b%C3%A6ck/building-a-robot-journalist-171554a6...
May have been discussed in the Algorithmic Journalism topic, but focus here more on the aspect of what role a "robot journalist" can play in relation to human journalists.

Report: Guide to Automated Journalism, Andreas Graefe (2016)
http://towcenter.org/research/guide-to-automated-journalism/
Recent review of the field. Very useful, both for its content and reference list as a source of further literature.

Popular article: How a robot wrote for Engadget, Souppouris (2016)
https://www.engadget.com/2016/08/15/robot-journalism-wordsmith-writer/
Article about use of Wordsmith in news production.

Article: Towards a sociology of computational and algorithmic journalism, C W Anderson (2012)
http://journals.sagepub.com/doi/abs/10.1177/1461444812465137
Some analysis of the role of automation in the world of conventional journalism. Don't get too drawn into the sociology!

Article: Decades of Automation in the Newsroom, Carl-Gustav Linden (2016)
http://www.tandfonline.com/doi/pdf/10.1080/21670811.2016.1160791
Highly relevant discussion of the history of automation in journalism and what role it has come to play, and may in future. Presents a contrast to the conventional take on the benefits and dangers of automated news production.

Article: Clarifying Journalism’s Quantitative Turn, Mark Coddington (2014)
http://www.tandfonline.com/doi/pdf/10.1080/21670811.2014.976400

Article: The Relevance of Algorithms, Tarleton Gillespie (2012)
http://iasc-culture.org/THR/channels/Infernal_Machine/wp-content/uploads...

Kurssin suorittaminen

During the first session, each participant will be allocated a topic (or one aspect of a topic) which they will investigate.

The second session will consist of a presentation by each participant about what they plan to present, based on an initial review of the literature. By this point, they will not have had time to do any detailed reading, but should have taken some time over the first week to look over the suggested reading and being investigation of the topic.

In each following session, one participant (or several) will give a presentation on their topic. We will then have a discussion of the material presented and the application of the technical content to news automation.

For each topic, a selection of references to relevant literature will be provided, and his will form the starting point for participants' investigation. They are not required to use all of this material and they should also search for further relevant literature.

Each participant should submit a short report summarising the literature they have read on their topic, together with a list of references to further literature they have found.