The latest from Nicolas J.S. Davies, author of an extremely well written Nimble book, Blood on Our Hands (2010):
In the first two parts of this report, I have estimated that about 2.4 million people have been killed as a result of the U.S. invasion of Iraq, while about 1.2 million have been killed in Afghanistan and Pakistan as a result of the U.S.-led war in Afghanistan. In the third and final part of this report, I will estimate how many people have been killed as a result of U.S. military and CIA interventions in Libya, Syria, Somalia and Yemen.
Source: How Many Millions Have Been Killed in America’s Post-9/11 Wars? Part 3: Libya, Syria, Somalia and Yemen – Consortiumnews
I am learning how to use TensorFlow.
One obvious use would be to classify book covers by sales categories.
In fact, I could probably charge for this.
I’ve finished updating the Nimble Books online bookstore and all 317 Nimble published books are now available once again.
This is referred to as the “Classic Bookstore” to distinguish from the “Algorithmic Bookstore”, which is coming soon.
Two limitations that I mean to remove in future:
- the store currently presents print and Kindle editions on separate pages.
- I have not yet added category & genre filters. So, if you are looking for a particular book, use the “search” bar.
I missed this when it first came out in 2015. A good article — understands what I was trying to achieve with PageKicker (and am once again!). Italian (Google Translate is available).
L’algoritmo nel libro
Orientato invece al “market research” è l’algoritmo di cui si avvale W.Frederick Zimmerman – tecnologo, avvocato ed ora anche editore – ideatore della piattaforma Nimble Books. Su questa piattaforma i sistemi AI lavorano soprattutto in funzione pre-pubblicazione, rastrellando in rete tutte le informazioni rilevanti per il progetto editoriale: l’originalità rispetto alle tendenze di mercato, l’adesione o meno dell’argomento con i contemporanei gusti del pubblico ecc. Insomma, per l’algoritmo di Zimmerman la priorità è comprendere – il più rapidamente possibile– quale potrebbe essere il “miglior libro potenziale del momento” da pubblicare prima che qualcun altro se ne accorga. È evidente, che il “core” di Nimble Books non è la scrittura automatica di un testo, ma la fase che la precede, ovvero quel processo in grado di fornire – dopo messo insieme tutti i pezzi giusti – all’editore la soluzione di un difficilissimo puzzle: veder finalmente apparire la sagoma di un best seller.
Aspetto interessante è che la soluzione scelta da Nimble Books propone un modello dove l’algoritmo non vuole sostituire tout court il fattore umano, ma opera a fianco ad esso integrandolo ed incrementandolo
Instead oriented to “market research” is the algorithm used by W.Frederick Zimmerman – technologist, lawyer and now also editor – creator of the Nimble Books platform. On this platform AI systems mainly work in pre-publication function, raking all the information relevant to the publishing project online: the originality with respect to market trends, the adhesion or not of the subject with the contemporary tastes of the public etc. . In short, for Zimmerman’s algorithm the priority is to understand – as quickly as possible – what could be the “best potential book of the moment” to be published before someone else realizes it. It is evident that the “core” of Nimble Books is not the automatic writing of a text, but the phase that precedes it, or that process that can provide – after put together all the right pieces – the editor the solution of a difficult puzzle: finally see the shape of a best seller.
Interestingly, the solution chosen by Nimble Books proposes a model where the algorithm does not want to replace the human factor tout court, but works alongside it, integrating it and increasing it.
Version 2.1.1 of PageKicker replaces an acronym-identifying regex with a narrower one that produces better results. It is still far from perfect.
#sed 's/[[:space:]]\+/\n/g' $txtinfile | sort -u | \
egrep '[[:upper:]].*[[:upper:]]' | sed 's/[\(\),]//g' | uniq
sed 's/[[:space:]]\+/\n/g' $txtinfile | sort -u | \
egrep [A-Z][a-zA-Z0-9+\.\&]*[A-Z0-9] | sed 's/[\(\),]//g' | uniq
I reviewed a number of text analytics approaches prior to selecting this simpler and stupider regex approach. Most of the available tools require that the full phrase be immediately followed by the acronym, often in parentheses. There’s one that doesn’t require that but it is in Java which means I’d have to traverse a learning curve to plug it in. Also, I’m not really looking just for acronyms, I’m also looking for technical initialisms such as B8 or B-8.
There is a very simple test file included in the commit that includes these terms:
Output from old version:
Output from new version:
After a year of leaving it fallow, I have had enough hobby time to update the open source version of PageKicker to version 2.0.1-Keats. The key new features are that it runs interchangeably on Linux and Mac versions of bash; replaces the (deprecated) Alchemy API Named Entity Recognizer with the well-known Stanford NER; and adds the booktype “draft-report”, which produces the first draft of a term-paper style report without the front and back matter of a book: a great way to jump-start a writing project.
Stay tuned for more about the significance of this for Nimble Books authors and readers.
Good news, the Nimble Books online store is operational once more. At present, there is just one book: WHY THE USS THRESHER (SSN-593) WAS LOST by Bruce Rule, but over the course of the next few months I will add the entire backlist (319 titles). In the meantime, if you are a pal, or a naval history enthusiast, or both, you can do me a big favor by ordering a copy — I need a few live tests to be sure things are working correctly!
Nimble Books is once again accepting new submissions for publication. Any nonfiction topic is welcome.
An author recently asked me some questions about the sales of his naval history book so I thought I would take the opportunity to summarize what I told him.
- Most naval history books sell between 250 and 1000 copies.
- Most books, including naval history books, sell at least half their lifetime sales in the first two years. Sales generally follow a pattern like 100, 50, 10, 5, 1 … Having steady sales is cause for pride that you have chosen a good topic and done a solid job.
- Only a few sell more than 10,000, and breakouts like Jim Hornfischer’s LAST STAND OF THE TIN CAN SAILORS are very rare.
- Battleships sell.
- Battlecruisers sell.
- Big battles sell.
- Carriers sell.
- Carrier aviation sells.
- Cruisers don’t sell.
- Destroyers, escort craft, and small combatants don’t sell.
- Merchantmen don’t sell.
- Submarines sell if they are spy subs, hunter-killers locked in a duel, rogue ICBM launchers, or tragically and mysteriously lost.
- SEALS sell, sell, sell, sell, sell.
The naval history reading audience is graying, and there hasn’t been a real naval battle since the Falklands, so it makes sense that the market is dwindling except for SEALs. The future prospects aren’t especially bright, ether, in my view. If there is a real no-holds-barred naval war in the near to mid-term, it will probably a) be unpleasant for the largely Western naval reading audience, as it will probably involve a lot of hypersonic anti-ship missiles and b) be accompanied by a nuclear exchange of some sort (see item a).